English course management system applied to English teaching

Abstract

The invention relates to the technical field of computer-aided teaching and intelligent data processing, and discloses an English course management system applied to English teaching, which comprises a topology construction module, a prefetching buffer module, a feature analysis module, a real-time pruning module and a path presentation module. According to the system, firstly, a semantic topology network is constructed based on course metadata, and subsequent nodes and attribute data of the subsequent nodes are pre-fetched at a background according to a topological relation to construct a candidate buffer pool; meanwhile, a feature analysis module analyzes the micro-interaction behavior of the user in real time to generate a feature vector; and the real-time pruning module calculates a cognitive damping coefficient according to the cognitive damping coefficient, and dynamically filters the buffer pool by utilizing a node retention formula in combination with difficulty increment and context switching cost value. And when a submission instruction is triggered, the system locks the optimal node in time to display the adaptive content, and corrects the topological weight by using the damping coefficient. According to the method, data retrieval and cognitive state calculation are parallelized, so that course real-time adaptation and low-delay interaction based on implicit feedback are realized.

Description

Technical Field

[0001] This invention relates to the field of computer-aided instruction and intelligent data processing technology, specifically to an English course management system applied to English teaching. Background Technology

[0002] With the popularization of computer-assisted language learning technology, intelligent English teaching systems have become important tools to assist users in language learning. Existing systems of this kind typically focus on providing personalized course content based on the user's learning progress in order to improve learning efficiency.

[0003] However, existing English teaching systems still have certain limitations in content recommendation and interaction mechanisms. Traditional systems rely heavily on users' explicit answer results (such as accuracy or scores) to determine their current language proficiency and formulate subsequent learning paths.

[0004] This result-based recommendation logic often ignores the user's micro-behavioral performance during the question-answering process. It is difficult to assess the user's real-time cognitive load and level of confusion through implicit data such as cursor movement trajectory and text editing frequency. As a result, the course content pushed by the system has problems such as too large a range of difficulty or awkward context connection, and cannot accurately adapt to the learner's instantaneous cognitive state.

[0005] Meanwhile, in terms of data processing architecture, traditional systems typically employ a serial data processing model, meaning that logical judgments and database queries only begin after receiving a user's submission command. This delayed data retrieval method can easily generate noticeable page loading delays in high-concurrency or large-data-volume scenarios, thereby disrupting the user's thought process and affecting the smoothness of the interaction.

[0006] Furthermore, existing systems' course resource organization is mostly based on preset static rules or fixed knowledge graphs, lacking an adaptive mechanism to dynamically adjust the correlation weights between knowledge points based on the actual interactive feedback of a large user group. This makes it difficult for the course path planning model to automatically optimize as user group cognitive patterns change, resulting in limited recommendation accuracy after long-term use. Summary of the Invention

[0007] To address the shortcomings of existing technologies, this invention provides an English course management system for English teaching, which solves the problems of existing technologies, such as reliance on explicit results leading to a lack of process awareness in content adaptation, serial data retrieval causing delays in interactive response, and the lack of a path weight adaptive optimization mechanism based on group feedback.

[0008] To achieve the above objectives, the present invention provides the following technical solution:

[0009] This invention provides an English course management system for English teaching, including a topology construction module, a prefetch buffer module, a feature parsing module, a real-time pruning module, and a path presentation module.

[0010] The topology construction module receives and processes externally input course metadata, generating discrete course nodes through instantiation and discretization. This module calculates the semantic distance between nodes and generates a context-switching cost value based on this distance, reflecting the degree of cognitive discontinuity caused by changes in context. Simultaneously, it incorporates difficulty weights to establish weighted connections between discrete course nodes. Finally, the discrete course nodes and weighted connections are mapped to a high-dimensional vector space to generate a semantic topology network.

[0011] The prefetch buffer module is used to obtain the current node identifier corresponding to the user's current learning position and locate the coordinates in the semantic topology network based on this identifier. Using a breadth-first search strategy, the prefetch buffer module retrieves the nodes adjacent to the current node to obtain the successor node pointers. Subsequently, based on the successor node pointers, the corresponding node attribute data is extracted from the graph database, specifically including the difficulty increment and context switching cost, and this data is loaded into memory to construct the candidate buffer pool.

[0012] The feature parsing module is used to monitor and parse the raw input signals generated by user-operated hardware. This module denoises and segments the raw input signals to generate a spatiotemporal operation sequence. Based on this, the module performs statistical analysis on the spatiotemporal operation sequence, normalizes and calculates the cumulative hover time and text length, and simultaneously calculates the modification entropy and modification frequency during the editing process. These are combined to form quantified behavioral indicators, and finally encapsulated to generate a standardized behavioral feature vector.

[0013] The real-time pruning module is used to dynamically filter the candidate buffer pool using real-time behavioral data. This module receives behavioral feature vectors and uses the cognitive damping formula to calculate a cognitive damping coefficient that reflects the user's current cognitive load by weighting cumulative hover time, text length, modification entropy, and modification frequency.

[0014] Subsequently, the real-time pruning module traverses the nodes in the candidate buffer pool, using the node retention formula combined with the calculated cognitive damping coefficient, the prefetched difficulty increment, and the cost of context switching to calculate the retention score for each node. This calculation process evaluates the degree of matching between the difficulty change of the candidate node and the user's current cognitive damping, and weights the cost of context switching. Finally, the module performs real-time filtering on the candidate buffer pool based on the retention score, removing node indices with scores below the dynamic pruning threshold, and generating a pruned buffer pool.

[0015] The path presentation module is responsible for executing the final decision and feedback. Responding to submission commands triggered by user actions, this module locks the pruned buffer pool upon triggering and determines the target successor node based on retention scores. It then retrieves data from the underlying database to display the appropriate course content. Simultaneously, this module extracts the cognitive damping coefficient generated by the interaction and feeds it back to the topology construction module. This is used to adjust the context-switching cost and difficulty weights of weighted edges in the semantic topology network, enabling adaptive updates to the system.

[0016] This invention provides an English course management system for English teaching. It has the following beneficial effects:

[0017] 1. This invention achieves dynamic adaptation of course content by quantifying micro-interaction behaviors. The system uses a feature parsing module to convert users' micro-operations such as cursor hovering and text modification into quantified behavioral feature vectors, and a real-time pruning module calculates the cognitive damping coefficient. Combined with a real-time filtering mechanism in the buffer pool, the system can filter out subsequent content of appropriate difficulty and coherent context based on the user's current cognitive load before the user submits an answer, avoiding the difficulty gap problem caused by the lack of process monitoring in traditional recommendation methods.

[0018] 2. This invention reduces system response latency through a parallel prefetching mechanism. While the user is learning the current node, the prefetching buffer module uses a topological neighborhood retrieval mechanism to preload the attribute data of potential successor nodes into memory, thus parallelizing the data query operation with the user's thinking process. When the user triggers the submission command, the system only needs to perform index locking and sorting in memory to complete the decision, thereby achieving fast page jump and content loading.

[0019] 3. This invention achieves adaptive optimization of the topology network through a parameter feedback mechanism. The path presentation module feeds back the cognitive damping coefficient generated during user interaction to the topology construction module, which is used to dynamically correct the weights of the connection edges in the semantic topology network. This enables the system to adjust the correlation strength between knowledge points based on the user's actual learning feedback, making the course path planning model more in line with the user's actual cognitive patterns. Attached Figure Description

[0020] Figure 1 This is an architecture diagram of an English course management system applied to English teaching according to the present invention;

[0021] Figure 2 This is a diagram of the feature parsing module architecture of the present invention;

[0022] Figure 3 This is a diagram of the architecture of the real-time pruning module of the present invention.

[0023] Among them, 100 is the topology construction module; 200 is the prefetch buffer module; 300 is the feature parsing module; 400 is the real-time pruning module; and 500 is the path presentation module. Detailed Implementation

[0024] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0025] Please see the appendix Figure 1 This invention provides an English course management system for English teaching, which mainly includes: a topology construction module 100, a prefetch buffer module 200, a feature parsing module 300, a real-time pruning module 400, and a path presentation module 500.

[0026] The topology building module 100 is used to establish the static data foundation of the system, transforming unstructured teaching resources into a computable weighted network. The topology building module 100 internally includes a node generation unit, an edge weighting unit, and a topology mapping unit.

[0027] The node generation unit receives course metadata from external input, which includes vocabulary, grammar rules, and dialogue fragments. The node generation unit performs instantiation and discretization processing on the course metadata to generate discrete course nodes in the graph database.

[0028] The edge-weighting unit receives discrete course nodes, calculates the semantic distance between any two nodes, and establishes weighted edges that include the cost of context switching and difficulty weights. The topology mapping unit integrates the discrete course nodes and weighted edges, mapping them to a high-dimensional vector space to generate a semantic topology network.

[0029] The prefetch buffer module 200 is connected to the topology building module 100 and is used to establish a forward-looking data cache in the background to provide the raw set to be processed for real-time filtering. The prefetch buffer module 200 internally includes a neighborhood retrieval unit and a buffer loading unit.

[0030] The neighborhood retrieval unit receives the current node identifier of the user and the semantic topology network generated by the topology construction module 100. The neighborhood retrieval unit locates the coordinates in the semantic topology network and uses a breadth-first search strategy to retrieve neighboring nodes, outputting successor node pointers. The buffer loading unit extracts corresponding attribute data from the database based on the successor node pointers. This attribute data includes the difficulty increment and the context switching cost. The buffer loading unit loads this data and constructs an initial candidate buffer pool, which is then sent to the real-time pruning module 400.

[0031] The feature parsing module 300 serves as the system's perception front end, transforming the underlying physical inputs into standardized behavioral features. Internally, the feature parsing module 300 includes a signal acquisition unit, an index quantization unit, and a vector encapsulation unit.

[0032] The signal acquisition unit monitors the raw input signals at the hardware level in real time, including cursor coordinate streams and keyboard key timestamps. The signal acquisition unit denoises and segments the raw input signals to generate a valid spatiotemporal operation sequence. The index quantification unit analyzes the spatiotemporal operation sequence, normalizes and calculates the cumulative hover time and text length, and simultaneously calculates the modification entropy and modification frequency during the editing process, combining them to form a quantified behavior index. The vector encapsulation unit encapsulates the quantified behavior index, generating a standardized behavior feature vector.

[0033] The real-time pruning module 400 is connected to the prefetch buffer module 200 and the feature parsing module 300, respectively, and serves as the core computing engine to dynamically filter the prefetched data in memory using real-time behavioral data. The real-time pruning module 400 internally includes a state calculation unit, a scoring calculation unit, and a dynamic filtering unit.

[0034] The state calculation unit receives the behavioral feature vector and calculates the user's current cognitive damping coefficient using the cognitive damping formula. The cognitive damping formula is as follows:

[0035] ;

[0036] In the formula: This represents the cognitive damping coefficient, which is the calculated value of the user's cognitive load at the current moment. This represents the first weighting coefficient, a preset constant used to adjust the effect of hovering time on the overall damping value; This indicates the cumulative hover time, which is the total duration the cursor stays within the key syntax area; Indicates the text length, i.e. the number of characters contained in the key syntax region, and is used as a normalization factor; This represents the second weighting coefficient, a preset constant used to adjust the impact of editing behavior on the overall damping value; This represents the summation operator, which sums the calculation results of all independently edited segments within the monitoring period; This represents the modification entropy, which is the degree of disorder in character changes during a single text editing operation. This indicates the modification frequency, which is the number of text deletion or rewriting operations that occur per unit of time.

[0037] The scoring calculation unit receives the cognitive damping coefficient and the candidate buffer pool, traverses the nodes in the candidate buffer pool, and calculates the retention score for each node using the node retention formula combined with the difficulty increment and the cost of context switching. The node retention formula is as follows:

[0038] ;

[0039] In the formula: This represents the retention score, which is the priority value for the candidate node to be retained in the buffer pool. The absolute value operator is used to take the non-negative value of the number inside the vertical bar. This represents the cognitive damping coefficient, taken from the calculation result of the cognitive damping formula; This represents the increment in difficulty, which is the numerical difference in the difficulty attribute between the candidate node and the current node. This represents a very small constant, a preset value used to prevent the denominator from being zero; This represents the penalty coefficient, a preset constant used to adjust the weight of the impact of context switching cost on retention score; This represents the cost of context switching, i.e., the cost incurred when jumping from the current node to a candidate node due to changes in context.

[0040] The dynamic filtering unit filters the candidate buffer pool based on the retention score, removes candidate node indices with scores below the threshold in real time, and generates a cleaned and pruned buffer pool.

[0041] The path presentation module 500 is connected to the real-time pruning module 400 and the topology construction module 100, and is used to execute the final decision, realize content switching and system parameter feedback updates. The path presentation module 500 internally includes an optimization decision unit, a content display unit and a parameter feedback unit.

[0042] The optimal decision-making unit listens for user submission commands. Upon triggering the submission command, it locks the pruned buffer pool and sorts and selects nodes based on retention scores to determine the target successor node. The content display unit retrieves data based on the index of the target successor node and displays the adapted course content on the interface.

[0043] The parameter feedback unit extracts the cognitive damping coefficient generated by this interaction and feeds it back to the topology construction module 100 to correct the edge weights in the topology network and complete the system update.

[0044] The topology construction module 100, serving as the system's foundational data architecture generator, is primarily responsible for transforming nonlinear English teaching resources into a computer-processable graph structure. This process is collaboratively completed by the node generation unit, the edge weighting unit, and the topology mapping unit.

[0045] The node generation unit is responsible for processing course metadata. Course metadata, as the raw input, encompasses vocabulary units (such as words and phrases), grammatical rules (such as tenses, voices, and sentence structures) and situational dialogue fragments (such as text from business meetings, airport boarding, and hospital visits). The node generation unit instantiates this course metadata, assigning a unique system identifier (UUID) to each individual teaching content and creating corresponding vertex objects in the graph database. Subsequently, the node generation unit discretizes these vertex objects, removing redundant formatting information from the original data and retaining only core semantic content and basic attribute tags (such as parts of speech, applicability levels, and keywords), thereby generating discrete course nodes. Discrete course nodes are the basic units that constitute the subsequent knowledge graph; they are independent in their storage structure and await subsequent logical connections.

[0046] The edge-weighting unit receives discrete course nodes output by the node generation unit and establishes logical connections, i.e., weighted edges, between any two nodes. The establishment of these edges is not based on a simple linear order, but rather on the semantic relevance and difficulty gradient between nodes. The edge-weighting unit calculates the semantic distance between nodes, which reflects the closeness of two teaching contents in terms of thematic context. For calculating semantic distance, those skilled in the art can use pre-trained natural language processing models (such as BERT or Word2Vec) to vectorize the node text and calculate the cosine similarity between vectors; this is a well-known technique in the field and will not be elaborated upon here.

[0047] Based on the calculated semantic distance, the edge-weighted unit generates the context switching cost ( The value of context switching (). The magnitude of this value relates to the degree of cognitive discontinuity that occurs when a user jumps from one knowledge point to another during the learning process due to changes in context. The cost of context switching ( The generated rules are based on preset semantic association rules. For example, when transitioning from a business meeting scenario to a business negotiation scenario, the semantic distance between the two is relatively close, and the generated rules are more relevant. The numerical value is relatively small (e.g., 0.2); however, the semantic span between a business meeting scenario and a biomedical scenario is large, resulting in... The value is relatively large (e.g., 0.8).

[0048] Simultaneously, the edge weighting unit also calculates the difficulty weight between nodes, which represents the numerical difference between two nodes in terms of language difficulty level (such as the CEFR standard). Finally, the edge weighting unit writes the context switching cost and difficulty weight into the attributes of the connecting edge, forming a weighted connecting edge.

[0049] The topology mapping unit is responsible for integrating discrete course nodes and weighted edges to construct a complete semantic topology network. The topology mapping unit maps nodes and edges in the graph database to a high-dimensional vector space, ensuring that each node contains not only its own static attributes but also its position information and neighborhood relationship features within the graph. This mapping process ensures that subsequent modules can quickly locate other nodes semantically related to the current node through vector retrieval. The generated semantic topology network, as the static data structure foundation, is fed to the prefetch buffer module 200 for subsequent dynamic path planning.

[0050] The prefetch buffer module 200 is primarily responsible for performing forward-looking data retrieval and cache construction using background computing resources while the user is learning the current course node, thereby avoiding database query latency during subsequent path switching. The data processing flow of this prefetch buffer module 200 is specifically executed by the neighborhood retrieval unit and the buffer loading unit.

[0051] The neighborhood retrieval unit receives the current node identifier of the user and the semantic topology network output by the topology construction module 100. The neighborhood retrieval unit first locates the coordinates of the current node in the high-dimensional vector space of the semantic topology network based on the current node identifier. After locking the coordinates, the neighborhood retrieval unit initiates a graph traversal algorithm, performing a breadth-first search (BFS) starting from the current node.

[0052] During the search process, the neighborhood retrieval unit searches only nodes that have a direct connection edge or a second-degree connection relationship with the current node, based on a preset retrieval depth threshold or retrieval quantity threshold.

[0053] The threshold for the number of nodes to be searched is set based on the system's memory allocation and real-time processing capabilities. For example, setting the threshold to 10 nodes ensures that the candidate set covers sufficient diversity without consuming excessive memory resources. The specific implementation logic of breadth-first search can be obtained from standard graph theory algorithm libraries, which are well-known technologies in the field and will not be elaborated upon here.

[0054] The neighborhood retrieval unit outputs the set of node index addresses found as successor node pointers, which point to the potential next-time teaching content in the database.

[0055] The buffer loading unit receives the successor node pointers output by the neighborhood retrieval unit and accesses the underlying database or memory-mapped file based on these pointers to extract the detailed attribute data of the corresponding node.

[0056] The buffer loading unit only extracts key metadata for subsequent decision calculations, mainly including the difficulty increment ( ) and the value of context switching ( Difficulty increment () The value is obtained by the buffer loading unit by calculating the difference between the difficulty attribute value of the candidate node and the difficulty attribute value of the current node. This value reflects the range of difficulty of the teaching content (positive value indicates increased difficulty, negative value indicates decreased difficulty).

[0057] The value of context switching ( The weight attribute is directly read from the connection edge in the semantic topology network, and the attribute value has been calculated during the topology construction phase.

[0058] After extracting the aforementioned key attributes, the buffer loading unit allocates independent storage space in the system's cache to construct a candidate buffer pool. The candidate buffer pool is a structured data list, where each record corresponds to a candidate node and contains the node's unique index, difficulty increment value, and context switching cost value. The buffer loading unit keeps the constructed candidate buffer pool in memory and maintains a waiting state for the real-time pruning module 400 to call and filter it.

[0059] This preloading mechanism ensures that when a user completes the current question and triggers the submission signal, the system does not need to perform time-consuming disk I / O operations, but instead processes the prepared data directly in memory, thus achieving rapid content transition.

[0060] The feature parsing module 300, serving as the system's real-time perception front-end, transforms the user's physical operational behavior into computer-understandable digital features, providing a quantitative basis for subsequent cognitive state calculations. The data processing logic of this feature parsing module 300 is executed sequentially by the signal acquisition unit, the index quantification unit, and the vector encapsulation unit.

[0061] The signal acquisition unit directly interfaces with the input / output interface of the underlying operating system, listening to and acquiring raw input signals in real time. These raw input signals contain all the micro-level operational data of the user on the human-computer interaction interface, specifically including the real-time coordinate stream of the cursor in the screen coordinate system. Key-down / key-up timestamps and character change events in text input boxes.

[0062] The signal acquisition unit performs noise reduction processing on the acquired raw input signal, using a moving average filter or a Kalman filter to smooth the cursor trajectory and eliminate invalid jitter caused by hand tremors or hardware sampling errors.

[0063] Meanwhile, the signal capture unit segments the signal according to the layout of interface elements (DOM structure), identifies and extracts the operation data that falls within the key grammar area (such as the input box of fill-in-the-blank questions, the keyword area of ​​reading comprehension), and generates a spatiotemporal operation sequence.

[0064] This spatiotemporal operation sequence not only records the trajectory of the operation but also marks the specific semantic region where the operation occurred. For the monitoring and coordinate mapping of the underlying input signals, those skilled in the art can use standard event-driven programming models (such as JavaScript Event Loop), which are well-known technologies in the field and will not be elaborated upon here.

[0065] The indicator quantification unit receives the cleaned spatiotemporal operation sequence and performs multi-dimensional statistical analysis on it to calculate and generate specific quantitative behavioral indicators. The indicator quantification unit first calculates the cumulative hovering time (…). The index is the sum of timestamp differences when the cursor coordinates fall inside the bounding box of a specific key syntax region. This index reflects the duration of user attention to the content in that region.

[0066] The indicator quantization unit simultaneously obtains the text length within this key syntax region. The number of words or characters contained in the statistical area is used as a normalization factor in subsequent calculations to eliminate the impact of different question lengths on dwell time.

[0067] In addition, the indicator quantification unit focuses on monitoring user behavior patterns during the text editing process and calculating modification entropy ( ) and modification frequency ( ).

[0068] Modify entropy ( The entropy is a numerical value calculated based on the concept of entropy in information theory, used to characterize the degree of disorder in a user's editing operations. For example, when a user repeatedly deletes and rewrites the same word (spell correction), the character changes are limited to a small edit distance, resulting in low editing entropy; while when a user completely erases the current answer multiple times and replaces it with a semantically completely different word (grammatical hesitation), the resulting editing entropy is high.

[0069] Modify frequency ( The count is the number of deletion (backspace / delete) or insertion operations that occur within a given time window (e.g., within the last 5 seconds). Combining these two metrics can effectively distinguish between spelling errors and cognitive confusion.

[0070] The vector encapsulation unit receives the aforementioned independent metrics output by the metric quantization unit, including cumulative hover time, text length, modification entropy, and modification frequency. The vector encapsulation unit performs normalization on these values, mapping them to a uniform numerical range (e.g., [0,1]) to eliminate calculation biases caused by different units of measurement. Subsequently, the vector encapsulation unit sorts the processed values ​​according to a preset dimensional order (e.g., [...]). The data are assembled into a behavioral feature vector. This behavioral feature vector, as a data carrier describing the user's current instantaneous implicit cognitive state, is synchronously transmitted to the real-time pruning module 400 in real time.

[0071] The real-time pruning module 400, as the core computing engine of the system, is used to establish a dynamic mapping relationship between the user's implicit cognitive state and the attributes of the course content, and to filter the pre-fetched candidate course content in real time through a parallel computing mechanism. The internal logic of the real-time pruning module 400 is executed sequentially by the state calculation unit, the scoring calculation unit, and the dynamic filtering unit.

[0072] The state calculation unit receives a behavior feature vector transmitted in real time by the feature parsing module 300. This vector contains standardized hover time, text length, modification entropy, and modification frequency data.

[0073] The state calculation unit uses a cognitive damping formula to weight and fuse the aforementioned multidimensional features to calculate the user's cognitive damping coefficient. This coefficient is a dynamically changing value used to quantify the intensity of the cognitive load faced by the user at the current learning moment. Specifically, the state calculation unit executes the following cognitive damping formula:

[0074] ;

[0075] In the formula: The cognitive damping coefficient is a calculated value representing the user's current level of cognitive load. The larger the value, the greater the user's current level of confusion or cognitive resistance. This represents the first weighting coefficient, a preset constant used to adjust the effect of hover time on the overall damping value. This constant is set based on the system's sensitivity to reading comprehension difficulties (e.g., setting...). =0.6); This indicates the cumulative hover time, which is the total duration the cursor stays within the key syntax area; The text length, i.e. the number of characters contained in the key grammatical region, is used as a normalization factor to eliminate the impact of differences in question length on the duration data. This represents the second weighting coefficient, a preset constant used to adjust the impact of editing behavior on the overall damping value. This constant is set based on the system's sensitivity to spelling / grammar output difficulties (e.g., setting...). =0.4); This represents the summation operator, which sums the calculation results of all independently edited segments within the monitoring period; This represents the modification entropy, which is the degree of disorder in character changes during a single text editing operation. This indicates the modification frequency, which is the number of text deletion or rewriting operations that occur per unit of time.

[0076] The scoring calculation unit receives the cognitive damping coefficient output by the state calculation unit. Simultaneously, the candidate buffer pool residing in the prefetch buffer module 200 memory is accessed. The scoring calculation unit traverses each candidate node in the buffer pool, reading the pre-loaded difficulty increment for that node ( ) and the value of context switching ( ).

[0077] The scoring calculation unit calculates the degree of match between each candidate node and the user's current cognitive state based on the node retention formula, generating a retention score. The core logic of this calculation process lies in selecting nodes whose difficulty changes are compatible with the user's current cognitive damping and whose context-switching costs are low. Specifically, the scoring calculation unit executes the following node retention formula:

[0078] ;

[0079] In the formula: The retention score represents the priority value of the candidate node in the buffer pool. The higher the score, the more suitable the node is as the teaching content for the next moment. The absolute value operator is used to take the non-negative value of the number inside the vertical bar. This represents the cognitive damping coefficient, taken from the calculation result of the cognitive damping formula; This represents the increment in difficulty, which is the numerical difference in the difficulty attribute between the candidate node and the current node. This represents a very small constant (e.g., 0.001) and is a preset value used to prevent the denominator from being zero. This represents the penalty coefficient, a preset constant used to adjust the weight of context switching costs on retention scores. This constant is set based on the degree of importance placed on coherence in the teaching strategy (e.g., setting...). =0.5); This represents the cost of context switching, i.e., the cost incurred due to the change in context when jumping from the current node to a candidate node. This item reflects the mismatch between the user's cognitive load and the adjustment of course difficulty. The smaller the mismatch, the higher the score of the first item.

[0080] The dynamic filtering unit performs real-time cleanup of the candidate pool based on the retention scores of each candidate node. The dynamic filtering unit sets a dynamic pruning threshold, which is based on the statistical characteristics of the score distribution in the current pool (such as the mean or median).

[0081] The dynamic filtering unit compares the retention score of each node with the pruning threshold. For nodes with scores below the threshold, the dynamic filtering unit immediately removes their index from memory, releases the resources they occupy, and generates a pruned buffer pool. This process continues during the intervals between user interactions, ensuring that when the user finally submits, the buffer pool only retains the candidate set that best matches their current cognitive state and contextual coherence requirements.

[0082] The path presentation module 500, as the final execution end of the system, is used to complete the zero-delay switching of teaching content based on the filtering results of the real-time pruning module 400, and is responsible for inputting the user's interaction status data back to the system's underlying layer to achieve adaptive system parameter updates. The operation flow of the path presentation module 500 is jointly executed by the optimization decision unit, the content display unit, and the parameter feedback unit.

[0083] The optimal decision-making unit is connected to the output interface of the real-time pruning module 400 and listens for submission command signals from the user interface.

[0084] The submission command signal is a system interrupt event triggered when the user clicks the "Submit Answer" or "Next Question" button. Upon detecting the submission command signal, the optimal decision-making unit immediately performs a locking operation on the pruned buffer pool residing in memory to prevent data write conflicts during the reading process.

[0085] The optimal decision-making unit reads the indexes of all remaining candidate nodes in the pruned buffer pool and their corresponding retention scores. The nodes are sorted in descending order (e.g., quicksort) based on their retention scores.

[0086] The optimal decision-making unit selects the node with the highest score in the sorted list as the target successor node. Since the nodes in the pruned buffer pool have been dynamically filtered multiple times by the real-time pruning module 400 during user interaction, the target successor node represents the node that best matches the user's cognitive state and contextual coherence at the moment of selection, without the need for additional database queries or logical judgments.

[0087] The content display unit receives the index information of the target successor node determined by the selection decision unit, and retrieves the corresponding entity data from the underlying teaching resource database based on the index, including text stems, multimedia audio files, and interactive component configurations. The content display unit uses a front-end rendering engine (such as a Web view renderer or a native UI framework) to load the retrieved data into the user interface, completing the display of the adapted course content.

[0088] Since the index information of the target successor node has been loaded into the cache during the prefetching buffer stage, the data retrieval process of the content display unit is actually a memory addressing operation, avoiding the disk I / O latency of traditional databases, thus ensuring that the interface jump after the user clicks submit has real-time responsiveness. Regarding the specific implementation technology of the front-end rendering engine, those skilled in the art can use existing HTML5 Canvas rendering or Android / iOS native view drawing technology, which are well-known technologies in the field and will not be elaborated upon here.

[0089] The parameter feedback unit is responsible for the closed-loop adaptive update of the system. The parameter feedback unit extracts the cognitive damping coefficient calculated by the state calculation unit during this interaction. The coefficient value is then sent as a feedback parameter to the topology building module 100.

[0090] The parameter feedback unit uses a cognitive damping coefficient to correct the weights of the connection edges between the current node and the target successor node in the semantic topology network. The specific correction logic is as follows: if the cognitive damping coefficient generated by this interaction is higher than a preset damping benchmark threshold (this benchmark threshold is set based on historical average cognitive load data, for example, 0.7), it indicates that the difficulty of the current path or the cost of context switching is underestimated by the system, and the parameter feedback unit will increase the context switching cost on that connection edge. The system adjusts the difficulty level by assigning a weight value to the network, or vice versa. Through this continuous parameter feedback mechanism, the system-generated corrected topology network can continuously evolve as user interaction data accumulates, making subsequent path planning more accurate.

[0091] The present invention provides an English course management system for English teaching, which maps discrete teaching resources into a semantic topology network containing semantic distance and difficulty gradient through a topology construction module 100. This changes the traditional linear teaching syllabus organization method and establishes the data foundation for nonlinear path planning.

[0092] While the user is at the current learning node, the prefetch buffer module 200 utilizes background computing resources to pre-build a candidate buffer pool containing potential successor nodes in memory based on topological neighborhood relationships, thus achieving forward-looking data preparation.

[0093] During this period, the feature parsing module 300 continuously monitors the user's micro-interaction behavior, transforming implicit operations such as cursor hovering and text modification into quantified behavioral feature vectors.

[0094] The real-time pruning module processes this real-time data in parallel, calculates the user's real-time cognitive damping state, and performs continuous negative feedback filtering on the candidate buffer pool residing in memory accordingly. The screening of candidate content is completed before the user triggers the submission action, eliminating nodes that do not meet the requirements of current cognitive load or contextual coherence.

[0095] Ultimately, the path presentation module 500 directly locks onto the optimal node in the pruned buffer pool, enabling rapid loading and display of course content, and corrects the weights of the underlying topology network through a parameter feedback mechanism. This system architecture, by parallelizing data retrieval and user interaction, achieves a shift from static rule-based push to dynamic path evolution based on implicit cognitive states, alleviating the technical problems of delayed teaching feedback and contextual disconnect in existing technologies.

Claims

1. An English course management system for English teaching, characterized in that, include: A topology building module (100) is used to receive and process course metadata from external input to build a semantic topology network; The prefetch buffer module (200) is used to obtain and retrieve the semantic topology network based on the current node identifier corresponding to the user's current learning position to obtain the successor node pointer, and extract the difficulty increment and context switching cost based on the successor node pointer to construct a candidate buffer pool. The feature parsing module (300) is used to listen to and parse the raw input signals generated by the user's operation of the hardware, and generate behavioral feature vectors; The real-time pruning module (400) is used to calculate the cognitive damping coefficient based on the behavioral feature vector using the cognitive damping formula, calculate the retention score of the nodes in the candidate buffer pool using the node retention formula in combination with the cognitive damping coefficient, and perform filtering on the candidate buffer pool based on the retention score to generate a pruned buffer pool. The path presentation module (500) is used to respond to the submission instruction signal triggered by the user submission operation, determine the target successor node from the pruned buffer pool to display the adapted course content, and use the cognitive damping coefficient to correct the semantic topology network.

2. The English course management system for English teaching according to claim 1, characterized in that, The topology building module (100) includes a node generation unit; The node generation unit is used to read the course metadata, perform instantiation and discretization processing on the course metadata, and generate discrete course nodes in the graph database.

3. The English course management system for English teaching according to claim 2, characterized in that, The topology construction module (100) also includes an edge weighting unit; The edge weighting unit is used to receive the discrete course nodes, calculate the semantic distance between the discrete course nodes, generate the context switching value based on the semantic distance, and establish a weighted connection edge that includes the context switching value and the difficulty weight.

4. The English course management system for English teaching according to claim 3, characterized in that, The topology building module (100) also includes a topology mapping unit; The topology mapping unit is used to integrate the discrete course nodes and the weighted connection edges, mapping the discrete course nodes and the weighted connection edges to a high-dimensional vector space to generate the semantic topology network.

5. An English course management system for English teaching according to claim 1, characterized in that, The prefetch buffer module (200) includes: The neighborhood retrieval unit is used to receive the current node identifier and the semantic topology network, locate the coordinates in the semantic topology network and retrieve neighboring nodes using a breadth-first search strategy, and output the successor node pointer. A buffer loading unit is used to extract the corresponding difficulty increment and context switching value from the graph database based on the successor node pointer, load the difficulty increment and context switching value, and construct the initial candidate buffer pool. The graph database is a database used to store the semantic topology network and its node attribute data.

6. The English course management system for English teaching according to claim 1, characterized in that, The feature parsing module (300) includes: The signal acquisition unit is used to monitor the original input signal in real time, perform noise reduction and segmentation processing on the original input signal, and generate a spatiotemporal operation sequence. The indicator quantification unit is used to analyze the spatiotemporal operation sequence, normalize and calculate the cumulative hover time and text length, and simultaneously statistically analyze the modification entropy and modification frequency during the editing process, combining them to form a quantitative behavior indicator. A vector encapsulation unit is used to encapsulate the quantified behavior indicators to generate the behavior feature vector.

7. An English course management system for English teaching according to claim 6, characterized in that, The real-time pruning module (400) includes a state calculation unit; The state calculation unit is used to receive the behavior feature vector and use the cognitive damping formula to perform weighted calculations on the cumulative hover time, the text length, the modification entropy, and the modification frequency to obtain the cognitive damping coefficient.

8. An English course management system for English teaching according to claim 7, characterized in that, The real-time pruning module (400) also includes a scoring calculation unit; The scoring calculation unit is used to traverse the nodes in the candidate buffer pool and calculate the retention score of each node by combining the node retention formula with the cognitive damping coefficient, the difficulty increment and the context switching cost.

9. An English course management system for English teaching according to claim 8, characterized in that, The real-time pruning module (400) also includes a dynamic filtering unit; The dynamic filtering unit is used to filter the candidate buffer pool based on the retention score, remove candidate node indices whose retention scores are lower than the dynamic pruning threshold, and generate the cleaned pruned buffer pool. The dynamic pruning threshold is preset based on the statistical characteristics of the score distribution in the candidate buffer pool.

10. An English course management system for English teaching according to claim 3, characterized in that, The path presentation module (500) includes: The optimal decision-making unit is used to listen to the submission instruction signal, lock the pruned buffer pool at the moment the submission instruction signal is triggered, and sort and select according to the retention score to determine the target successor node; The content display unit is used to retrieve data based on the index of the target successor node and display the adapted course content on the interface. The parameter feedback unit is used to extract the cognitive damping coefficient generated by this interaction and feed it back to the topology construction module (100) to correct the context switching cost and difficulty weight of the weighted connection edge in the semantic topology network, thereby completing the system update.

Images