An intent-driven personalized learning path recommendation method

Abstract

The application discloses an intention-driven personalized learning path recommendation method, and belongs to the recommendation field. First, a benchmark learning path conforming to knowledge prerequisite logic is generated based on a current knowledge state portrait of a user and a pre-constructed knowledge graph. Then, in the process that the user executes the benchmark learning path, the active interaction behavior of the user is monitored in real time, the learning intention of the user to a current learning knowledge point is identified through a natural language processing technology, if the identified learning intention matches a preset trigger condition, a local review path is generated at the current learning knowledge point based on the knowledge graph and the current knowledge state portrait, and the user is guided to learn; after the local review path is executed, the benchmark learning path is returned to continue learning. The application combines static learning path planning and dynamic intention response, improves the accuracy and efficiency of learning intervention, and optimizes the learning experience and effect.

Description

Technical Field

[0001] This invention belongs to the field of recommendation, and in particular relates to an intent-driven personalized learning path recommendation method. Background Technology

[0002] Personalized learning recommendation is one of the core research directions in modern educational technology. It aims to provide learners with customized learning content and sequences by analyzing individual differences. In this field, building learning paths based on knowledge graphs is a mainstream technical approach. Knowledge graphs can model course knowledge in a graph structure, representing knowledge points as nodes, and describing logical dependencies such as prerequisites and relevance between knowledge points through directed edges, thus laying the foundation for learning path planning. Simultaneously, by analyzing users' historical learning behavior data, a current knowledge state profile reflecting their mastery of each knowledge point can be constructed, providing a basis for personalized recommendations.

[0003] Existing personalized learning path recommendation methods typically employ a static "one-time planning" model. The goal of this path planning is to master all weak knowledge points and form an optimal linear learning sequence from the current state to the target state. After the path is generated, the user learns according to this fixed sequence. The system updates the user's state mainly by evaluating passive behaviors such as answering questions and periodically adjusting the knowledge state profile.

[0004] However, because the learning process is dynamic, users generate immediate and proactive learning needs while learning along a preset path. For example, they may want a deeper understanding of a particular knowledge point or feel confused and need to solidify their foundation. These dynamic learning needs, stemming from the user's genuine intent, cannot be perceived or responded to by static path planning methods. Furthermore, the system's assessment of the user's knowledge status has a certain lag and inaccuracy, leading to untimely path adjustments. Therefore, existing technical solutions struggle to respond in real-time to the user's proactive, intent-driven learning needs during the execution of the learning path, resulting in insufficient flexibility of the learning path, a disconnect from the user's real-time state, and ultimately impacting learning efficiency and experience. Summary of the Invention

[0005] To address the aforementioned issues, this invention proposes an intent-driven personalized learning path recommendation method, which generates a personalized learning path that can be dynamically adjusted based on the user's real-time learning intent.

[0006] The technical solution adopted in this invention is as follows:

[0007] Firstly, this invention proposes an intent-driven personalized learning path recommendation method, comprising the following steps:

[0008] Step 1: Based on the user's current knowledge status profile and the pre-built knowledge graph, generate a baseline learning path that conforms to the knowledge pre-learning logic; the current knowledge status profile records the user's mastery status of each knowledge point, including mastered, mastering, and not mastered.

[0009] Step 2: During the process of the user executing the baseline learning path, monitor the user's active interaction behavior in real time, and identify the user's learning intention for the current knowledge point through natural language processing technology. The learning intention includes conceptual questions, basic consolidation and in-depth understanding.

[0010] If the identified learning intention matches the preset triggering conditions, a local review path is generated at the current learning knowledge point based on the knowledge graph and the current knowledge state profile, and the user is guided to learn; after the local review path is completed, the user returns to the base learning path to continue learning.

[0011] The local review path is generated based on the knowledge graph, which defines a network of related knowledge points originating from the currently learned knowledge points, and is combined with the current knowledge state profile.

[0012] The triggering condition is: the learning intention of the current learning knowledge point is basic consolidation, and the mastery status of the current learning knowledge point in the current knowledge status profile is "under mastery".

[0013] Furthermore, it also includes the update steps for the benchmark learning path:

[0014] The current knowledge status profile is updated in real time or periodically based on the user's learning behavior data;

[0015] When a preset update trigger condition is met, a new baseline learning path is regenerated based on the updated knowledge state profile, replacing the original baseline learning path; the update trigger condition includes at least one of the following:

[0016] The user has completed learning the preset number of knowledge points;

[0017] Reaching the preset path update cycle time point;

[0018] In the user's knowledge status profile, the mastery status changes and the magnitude of the change exceeds a preset threshold.

[0019] Furthermore, the process of generating the benchmark learning path includes:

[0020] The graph traversal begins with a user-specified or system-randomized starting knowledge point, ensuring that the generated path conforms to the preset path length or learning time constraints while satisfying the prerequisite logic.

[0021] During graph traversal, knowledge points whose mastery status is "not mastered" or "mastering" are selected from the set of neighboring knowledge points of the current knowledge point as candidates for the next knowledge point.

[0022] For candidate knowledge points, check whether all prerequisite knowledge points have been mastered. If they have been mastered, add them to the path. When there are multiple knowledge points that meet the conditions, select the knowledge point with the lowest mastery probability in the current knowledge status profile and add it to the path.

[0023] If the prerequisite knowledge points for a candidate knowledge point are not fully mastered, then select the core knowledge points from the prerequisite knowledge points that have not been mastered or are currently mastered and add them to the path.

[0024] Furthermore, the core knowledge points are knowledge points determined by calculating the network centrality index of the knowledge graph, and the network centrality index includes at least one of degree centrality, betweenness centrality, and proximity centrality.

[0025] Furthermore, the process of identifying learning intentions includes:

[0026] Collect user's proactive interaction data, which includes behavior type, behavior timestamp, and target knowledge point. The behavior type is divided into notes or questions by keyword matching or sentence structure analysis of the interaction content through natural language processing technology.

[0027] Based on the behavior type, frequency, and time decay factor of the active interaction behavior, the intensity of the student's intention for the current learning knowledge point is calculated. The intensity of the intention is the weighted value of the note frequency, question frequency, and time decay factor per unit time.

[0028] Based on the relationship between the intensity of intent and a preset threshold, learning intent is categorized into conceptual questions, basic reinforcement, or in-depth understanding.

[0029] Furthermore, the process of generating local review paths includes:

[0030] Extract the dependency chain and the relevance chain of the currently learned knowledge point from the knowledge graph. The dependency chain contains the prerequisite knowledge points and subsequent knowledge points of the currently learned knowledge point, and the relevance chain contains related knowledge points with a similarity to the currently learned knowledge point that is higher than a threshold.

[0031] From the prior knowledge points and related knowledge points, select the knowledge points whose mastery status is "not mastered" or "mastering" in the current knowledge status profile, and use them as knowledge points to be reviewed;

[0032] Calculate the recommendation strength of the knowledge points to be reviewed, select the knowledge points with the highest recommendation strength as the review focus, and generate a local review path with a customizable length.

[0033] Furthermore, if the review focus belongs to the dependency chain of the knowledge point to be reviewed, the generated local review path shall at least include the review focus itself, and may selectively include its prerequisite knowledge points and / or subsequent knowledge points on the dependency chain, and the order of knowledge points in the path shall conform to the prerequisite logic of the knowledge graph.

[0034] If it belongs to the relevance chain of the knowledge points to be reviewed, the generated local review path includes the review focus and one or more related knowledge points with high recommendation intensity.

[0035] If they belong to two different relation chains, then a generation method is randomly determined.

[0036] Furthermore, the formula for calculating the recommendation strength is as follows:

[0037] ;

[0038] in, K represents the knowledge points to be reviewed, and K represents the knowledge points currently being studied. These are the knowledge points to be reviewed. betweenness centrality, These are the knowledge points to be reviewed. The distance to the current learning knowledge point K, The user's attitude towards reviewing knowledge points is derived from the current knowledge status profile. The probability of mastering These are the knowledge points to be reviewed. Recommended strength.

[0039] In a second aspect, the present invention provides an electronic device, including a memory and a processor;

[0040] The memory is used to store computer programs;

[0041] The processor is configured to implement the intent-driven personalized learning path recommendation method as described above when executing the computer program.

[0042] Thirdly, the present invention proposes a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the intent-driven personalized learning path recommendation method as described above.

[0043] The beneficial effects of this invention are:

[0044] This invention combines static learning path planning with dynamic intent response to achieve intelligent and personalized learning guidance. By monitoring and analyzing users' proactive interactive behaviors in real time, it accurately identifies students' immediate and intrinsic learning intentions and inserts a local review path when trigger conditions are met. Furthermore, the local review path introduced in this invention is not simply a repetition or jump to knowledge points, but rather dynamically generates a personalized path based on the knowledge graph's relational network, using the current learning knowledge point as the origin. This path focuses on the current problem point itself and can also connect with its prerequisite or subsequent knowledge points for knowledge consolidation, and all operations are completed under the premise of strictly adhering to knowledge logic. This invention improves the accuracy and efficiency of learning intervention, effectively avoids the time wasted by blind review, and optimizes the learning experience and results. Attached Figure Description

[0045] Figure 1 This is a flowchart illustrating an intent-driven personalized learning path recommendation method proposed in this invention.

[0046] Figure 2 This is a flowchart illustrating the process of generating a benchmark learning path;

[0047] Figure 3 This is a flowchart illustrating the process of identifying learning intentions;

[0048] Figure 4 This is a flowchart illustrating the process of generating a local review path;

[0049] Figure 5 This is a schematic diagram of the electronic device terminal structure of the present invention. Detailed Implementation

[0050] The present invention will be further described and illustrated below with reference to specific embodiments. The embodiments described are merely examples of the content of this disclosure and do not limit the scope of the invention. The technical features of each embodiment in the present invention can be combined accordingly, provided that there is no mutual conflict.

[0051] The accompanying drawings are merely illustrative of the invention and are not necessarily drawn to scale. Some of the block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. These functional entities can be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor devices and / or microcontroller devices.

[0052] The flowchart shown in the attached diagram is merely an illustrative example and does not necessarily include all steps. For example, some steps may be broken down, while others may be combined or partially combined; therefore, the actual execution order may change depending on the specific circumstances.

[0053] like Figure 1 As shown, the present invention proposes an intent-driven personalized learning path recommendation method, which includes the following steps:

[0054] S1 generates a baseline learning path that conforms to the logic of prior knowledge acquisition, based on the user's current knowledge state profile and a pre-built knowledge graph.

[0055] This step is based on an objective assessment of the student's current knowledge status. Using the knowledge logic relationships described by the knowledge graph, an optimal learning sequence from the current state to the target state is planned. The goal is to master all weak knowledge points, and this optimal learning sequence serves as the baseline learning path.

[0056] It's important to note that knowledge graphs are a technique for modeling and representing course knowledge using graph structures. They express knowledge points in a course as triples of "node-edge-node," where a node represents a knowledge point, such as "quadratic equation" or "Pythagorean theorem." Edges represent prerequisites and related relationships between knowledge points. Prerequisite relationships are represented by directed edges; for example, A→B indicates that knowledge point A must be learned before knowledge point B can be learned.

[0057] The user's current knowledge status profile is a probability estimate (or mastery probability) of their mastery level for each knowledge point, calculated by the system using a machine learning model based on the student's historical learning behavior data. This process can be implemented using existing technologies, and the mastery level of knowledge points is updated based on the student's continuously generated new learning behavior data. This profile is for each knowledge point k. i A three-color labeling system is used to indicate the status of the knowledge point. Generally, red indicates not mastered, meaning that the student's probability of mastering the knowledge point is below the lower threshold; yellow indicates mastering, meaning that the student's probability of mastering the knowledge point is between the lower and higher thresholds; and green indicates mastering, meaning that the student's probability of mastering the knowledge point is above the higher threshold.

[0058] In this embodiment, as Figure 2 As shown, a preferred implementation process is as follows:

[0059] (1.1) Select target knowledge points.

[0060] The knowledge points that need to be learned or consolidated are selected from the knowledge status profile as target knowledge points (marked in red and yellow). Knowledge points marked in green will be skipped in the baseline path to save learning time and improve efficiency.

[0061] (1.2) Knowledge point topology sorting based on knowledge graph.

[0062] Starting with a randomly selected knowledge point (which can be red, green, or yellow, as specified by the student), a graph traversal algorithm, such as Breadth-First Search (BFS) or Depth-First Search (DFS), is used to find knowledge points that need to be learned or reinforced, while respecting the dependencies of directed edges. For example, let C(k) represent the color of knowledge point k. Starting from k, traverse its neighboring knowledge points (connected by directed or undirected edges), prioritizing knowledge points with the colors red or yellow as candidates. For each candidate knowledge point, check if all its prerequisites (connected by directed edges pointing to it) have been mastered. If so, the knowledge point is recommended. If there are multiple candidates, they are ranked according to their mastery probability, with the knowledge point with the lowest mastery probability being learned first. If a candidate knowledge point does not meet its prerequisites, the unmastered prerequisites are recommended to be learned first, and the most important core knowledge point is selected from the unmastered prerequisites.

[0063] In this embodiment, the knowledge graph is marked with core knowledge points. The core knowledge points are obtained by calculating the network centrality index of each knowledge point in the knowledge graph. The network centrality index includes at least one of degree centrality, betweenness centrality, and proximity centrality.

[0064] When a candidate knowledge point meets the prerequisites, assuming the current knowledge point in the planned path is k, let N(k) represent the neighbor set of k, and P(k) represent the prerequisite set of k, then the next knowledge point k recommended by the path is... next for:

[0065]

[0066] For example: The target knowledge points to be learned or reinforced are A, B, and C, with edges A→B and A→C (A is a prerequisite for B and C). Starting from A, if C(A) = green, then skip the subsequent content of A and check B and C. Assuming C(B) = red, C(C) = yellow, and the prerequisite A for B and C has been mastered, then B is recommended first (because red indicates that it has not been mastered and needs more learning).

[0067] (1.3) Output the baseline learning path.

[0068] Based on the predetermined learning duration or other known learning objectives, generate a linear sequence of knowledge points L = [k1, k2, ..., k n ], such that any directed edge k i → k j In sequence L, k i All appear in k j Previously, the relationship of prior treatment was observed.

[0069] S2, during the process of the user executing the baseline learning path, monitor the user's active interaction behavior in real time, identify the user's learning intention for the current knowledge point through natural language processing technology, and determine whether to trigger local path modification based on the learning intention.

[0070] This step involves analyzing students' proactive interaction behavior in real time as they follow the baseline learning path to gain insight into their true learning intentions.

[0071] In this embodiment, as Figure 3 As shown, a preferred implementation process is as follows:

[0072] (2.1) Real-time monitoring and behavioral data collection.

[0073] As students learn through the terminal tool following the recommended learning path, they can select learning content by making various selections, such as marking learning content (text, formulas), adding text or voice notes, asking questions via voice or text, or clicking specific function buttons.

[0074] The core task of natural language processing for behavioral text content is to classify the selected content and determine whether the student's behavior is a note-taking exercise or a question. For example, if the text contains keywords or interrogative sentences such as "why," "don't understand," or "how to deduce," it may be identified as a question; if it contains words such as "summary," "key points," or "common mistakes," it may be identified as a note-taking exercise.

[0075] By collecting students' active interactive behaviors, each behavior is encapsulated into a data unit B={b1,b2,...,b...} n}, each action b i This includes, but is not limited to, behavior type (notes or questions), behavior timestamp, and target knowledge point, where the target knowledge point refers to the specific knowledge point that is currently being learned.

[0076] (2.2) Intent recognition.

[0077] Define an intent intensity function to quantify students' learning needs for a specific knowledge point K, as follows:

[0078] For knowledge point K, calculate its intention intensity I(K), which quantifies the student's self-evaluation of knowledge point K into a probability value:

[0079]

[0080] in, It represents the frequency of note-taking per unit of time, reflecting learning activity. T is the length of the time window. It represents the number of times knowledge point K is noted within the time window T. The frequency of questions asked about knowledge point K reflects the level of difficulty in understanding. It refers to the number of questions asked about knowledge point K. It refers to the number of notes taken for knowledge point K. It is the time decay factor (the smaller the time decay, the more urgent the demand). This is the attenuation coefficient, used to control the attenuation rate affected by time (initial value 0.1), t current It is the current time, t last This is the time of the most recent operation on knowledge point K. It is the weighting coefficient.

[0081] The intent type can be identified as deep understanding, basic reinforcement, or conceptual questioning based on the intent intensity I(K) of the behavior and a preset threshold. The expression for intent recognition is:

[0082]

[0083] in, and They are a higher threshold and a lower threshold, respectively.

[0084] During the student's learning process, each knowledge point in the baseline learning path gradually changes from red to green, indicating progress from not mastering it to mastering it. This process is updated based on the system's automatically assessed probability of mastery. However, system assessments typically involve probabilistic modeling of discrete behavioral data, which can introduce lag or inaccuracy, leading to errors. Therefore, this step monitors the student's intent regarding the current knowledge point, using this intent to drive path optimization. If the student's intent is to consolidate basic knowledge and the current knowledge point is in yellow, it determines that a local review path needs to be inserted at that point. This local review path recommends prior knowledge review and basic practice questions, triggering the next step.

[0085] S3, at a given knowledge point, based on the knowledge graph and the current knowledge state profile, generate a local review path and guide the user to learn; after the local review path is completed, return to the baseline learning path to continue learning.

[0086] Specifically, starting from a given knowledge point K, which is the knowledge point that triggers the local review path in step S2, we find knowledge points that have a dependency or similarity relationship with the current knowledge point K and whose status is red or yellow as the knowledge points to be reviewed, and use them to construct the local review path.

[0087] In one specific embodiment of the present invention, such as Figure 4 As shown, the process of generating a local review path includes:

[0088] (3.1) Extract the dependency and relevance chains of knowledge point K from the knowledge graph:

[0089]

[0090]

[0091] in, It is a dependency relationship chain. It is a correlation relationship chain. It is the m-th prerequisite knowledge point for knowledge point K. It is the nth subsequent knowledge point of knowledge point K. These are knowledge points related to knowledge point K. It refers to the similarity between knowledge points.

[0092] (3.2) Select the knowledge points that are in red or yellow in the current knowledge status profile from the prerequisite knowledge points (excluding subsequent knowledge points) or the related knowledge points in the dependency relationship chain as knowledge points to be reviewed.

[0093] The recommended strength of the knowledge points to be reviewed can be calculated using the following formula:

[0094]

[0095] in, These are knowledge points that need to be reviewed; consider them as weak knowledge points that are relatively close to the current knowledge point K. It is a knowledge point betweenness centrality, It is a knowledge point Distance to K The knowledge points are automatically evaluated by the system. The probability of mastering something can be obtained through a user's current knowledge status profile. It refers to the recommendation intensity, selecting the knowledge points with the highest recommendation intensity as the focus of the local review path.

[0096] (3.3) Based on the type of relationship chain to which the review focus belongs, different strategies are used to generate local review paths.

[0097] In this step, the selected knowledge points to be reviewed are identified. The review focus and its position in the relation chain are considered. If it belongs to a dependency relation chain, the generated local review path must at least include the review focus itself and may selectively include its prerequisite knowledge points and / or subsequent knowledge points in the dependency relation chain. In the final generated path, the learning order of all knowledge points must conform to the prerequisite logic defined by the knowledge graph. For example, if a student is currently learning knowledge point K, "Pythagorean theorem," and the review focus is "properties of right triangles," and this review focus belongs to a knowledge point in the dependency relation chain, then the system will generate a short review path that directly targets this weak knowledge point, or combine the review focus and its prerequisite knowledge point "square operations" and subsequent knowledge point "simple applications of the Pythagorean theorem" in the order of the dependency relation chain into a single local review path.

[0098] If it belongs to a relevance chain, the generated local review path includes the review focus and one or more related knowledge points with high recommendation strength. These knowledge points can be studied in order of recommendation strength. For example, if a student is currently learning knowledge point K, "Pythagorean theorem," and the review focus is "properties of right triangles," which belongs to a knowledge point in a relevance chain, then a related knowledge point with high recommendation strength, "difference of squares formula," will be selected from the relevance chain, and these two related knowledge points will be combined into a local review path in the order of recommendation strength: "properties of right triangles" and "difference of squares formula."

[0099] The length of a local review path can be customized. After the local review path is completed, the system returns to the baseline learning path. If the path belongs to two different relation chains, the path is randomly determined, or the generation method of the dependency relation chain is preferred.

[0100] In one specific embodiment of the present invention, since the user's continuous learning behavior will affect the user's knowledge state profile, a baseline learning path can be regenerated using the updated knowledge state profile according to the system's preset rules to meet the user's learning needs. The above description only focuses on a single baseline learning path, and those skilled in the art can update and optimize it based on this.

[0101] An alternative benchmark learning path has the following update steps:

[0102] The system updates the current knowledge status profile in real-time or periodically based on user learning behavior data. For example, the system collects user learning behavior data at fixed time intervals. This behavior data includes, but is not limited to: answer accuracy, answer time, video viewing completion rate, number of interactions, and note content. By modeling and analyzing this behavior data, the system calculates and adjusts the mastery probability value of each knowledge point in the profile. For example, if a user answers practice questions related to a certain knowledge point correctly consecutively, their mastery probability for that knowledge point will be increased; conversely, if they answer incorrectly consecutively or show hesitation, the mastery probability will be decreased. In this way, the knowledge status profile reflects the user's latest and most accurate learning status.

[0103] When a preset update trigger condition is met, a new baseline learning path is regenerated based on the updated knowledge state profile, replacing the original baseline learning path; the update trigger condition includes at least one of the following:

[0104] The user has completed a preset number of knowledge points. For example, the system can be set to trigger a global path update every 10 knowledge points completed. This avoids the user being confused by overly frequent path replanning and ensures that after completing a meaningful learning stage, the system can optimize the learning path based on the latest knowledge status, so that the learning path for the next stage is more in line with the user's current level.

[0105] The system can automatically update the learning path for the following week every Monday, ensuring the continuity and predictability of learning progress.

[0106] In a user's knowledge profile, the mastery level changes and the magnitude of the change exceeds a preset threshold. This method is not dependent on a fixed quantity or time. For example, if a user suddenly masters several weak knowledge points through one or more efficient local reviews, this sudden progress will lead to a significant change in their knowledge structure. At this point, even if the fixed quantity or time update conditions are not met, the system can immediately trigger a path update, adjusting the new learning focus to other knowledge points that still need improvement, preventing the user from wasting time on content they have already mastered, thereby maximizing learning efficiency.

[0107] It should be noted that the intent-driven personalized learning path recommendation method in the above embodiments can essentially be executed by a computer program. Therefore, similarly, based on the same inventive concept, another preferred embodiment of the present invention also provides a computer electronic device corresponding to the method provided in the above embodiments, which includes a memory and a processor;

[0108] The memory is used to store computer programs;

[0109] The processor is configured to implement the intent-driven personalized learning path recommendation method in the above embodiments when executing the computer program.

[0110] From a hardware perspective, such as Figure 5 The diagram shown is a hardware structure diagram provided in this embodiment. In addition to the processor, memory, network interface and non-volatile memory shown in the diagram, any device with data processing capabilities in the embodiment may also include other hardware depending on the actual function of the device with data processing capabilities, which will not be described in detail here.

[0111] When the logical instructions in the aforementioned memory can be implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product, which is stored in a storage medium.

[0112] Therefore, based on the same inventive concept, another preferred embodiment of the present invention also provides a computer-readable storage medium corresponding to the method provided in the above embodiments. The storage medium stores a computer program that, when executed by a processor, can realize the intent-driven personalized learning path recommendation method in the above embodiments.

[0113] It is understood that the computer-readable storage medium can be an internal storage unit of any data processing device described in any of the foregoing embodiments, such as a hard disk or memory. The computer-readable storage medium can also be an external storage device of any data processing device, such as a plug-in hard disk, smart media card (SMC), SD card, flash card, etc., equipped on the device. Furthermore, the computer-readable storage medium can include both internal storage units and external storage devices of any data processing device. The computer-readable storage medium is used to store the computer program and other programs and data required by the data processing device, and can also be used to temporarily store data that has been output or will be output.

[0114] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the invention. Those skilled in the art can make various changes and modifications without departing from the spirit and scope of the invention. Therefore, all technical solutions obtained through equivalent substitution or transformation fall within the protection scope of the present invention.

Claims

1. An intent-driven personalized learning path recommendation method, characterized in that, Includes the following steps: Step 1: Based on the user's current knowledge status profile and the pre-built knowledge graph, generate a baseline learning path that conforms to the knowledge pre-learning logic; the current knowledge status profile records the user's mastery status of each knowledge point, including mastered, mastering, and not mastered. Step 2: During the process of the user executing the baseline learning path, monitor the user's active interaction behavior in real time, and identify the user's learning intention for the current knowledge point through natural language processing technology. The learning intention includes conceptual questions, basic consolidation and in-depth understanding. If the identified learning intention matches the preset triggering conditions, a local review path is generated at the current learning knowledge point based on the knowledge graph and the current knowledge state profile, and the user is guided to learn. After the local review path is completed, return to the baseline learning path to continue learning; The local review path is generated based on the knowledge graph, which defines a network of related knowledge points originating from the currently learned knowledge points, and is combined with the current knowledge state profile. The triggering condition is: the learning intention of the current learning knowledge point is basic consolidation, and the mastery status of the current learning knowledge point in the current knowledge status profile is "under mastery".

2. The intention-driven personalized learning path recommendation method according to claim 1, characterized in that, It also includes the update steps for the benchmark learning path: The current knowledge status profile is updated in real time or periodically based on the user's learning behavior data; When a preset update trigger condition is met, a new baseline learning path is regenerated based on the updated knowledge state profile, replacing the original baseline learning path; the update trigger condition is any one of the following: The user has completed learning the preset number of knowledge points; Reaching the preset path update cycle time point; In the user's knowledge status profile, the mastery status changes and the magnitude of the change exceeds a preset threshold.

3. The intention-driven personalized learning path recommendation method according to claim 1, characterized in that, The process of generating the benchmark learning path includes: The graph traversal begins with a user-specified or system-randomized starting knowledge point, ensuring that the generated path conforms to the preset path length or learning time constraints while satisfying the prerequisite logic. During graph traversal, knowledge points whose mastery status is "not mastered" or "mastering" are selected from the set of neighboring knowledge points of the current knowledge point as candidates for the next knowledge point. For candidate knowledge points, check whether all prerequisite knowledge points have been mastered. If they have been mastered, add them to the path. When there are multiple knowledge points that meet the conditions, select the knowledge point with the lowest mastery probability in the current knowledge status profile and add it to the path. If the prerequisite knowledge points for a candidate knowledge point are not fully mastered, then select the core knowledge points from the prerequisite knowledge points that have not been mastered or are currently mastered and add them to the path.

4. The intention-driven personalized learning path recommendation method according to claim 3, characterized in that, The core knowledge points are determined by calculating the network centrality index of the knowledge graph, which includes at least one of degree centrality, betweenness centrality, and proximity centrality.

5. The intention-driven personalized learning path recommendation method according to claim 1, characterized in that, The process of identifying learning intentions includes: Collect user's proactive interaction data, which includes behavior type, behavior timestamp, and target knowledge point. The behavior type is divided into notes or questions by keyword matching or sentence structure analysis of the interaction content through natural language processing technology. Based on the behavior type, frequency, and time decay factor of the active interaction behavior, the intensity of the student's intention for the current learning knowledge point is calculated. The intensity of the intention is the weighted value of the note frequency, question frequency, and time decay factor per unit time. Based on the relationship between the intensity of intent and a preset threshold, learning intent is categorized into conceptual questions, basic reinforcement, or in-depth understanding.

6. The intention-driven personalized learning path recommendation method according to claim 1, characterized in that, The process of generating a local review path includes: Extract the dependency chain and the relevance chain of the currently learned knowledge point from the knowledge graph. The dependency chain contains the prerequisite knowledge points and subsequent knowledge points of the currently learned knowledge point, and the relevance chain contains related knowledge points with a similarity to the currently learned knowledge point that is higher than a threshold. From the prior knowledge points and related knowledge points, select the knowledge points whose mastery status is "not mastered" or "mastering" in the current knowledge status profile, and use them as knowledge points to be reviewed; Calculate the recommendation strength of the knowledge points to be reviewed, select the knowledge points with the highest recommendation strength as the review focus, and generate a local review path with a customizable length.

7. The intention-driven personalized learning path recommendation method according to claim 6, characterized in that, If the review focus belongs to the dependency chain of the knowledge point to be reviewed, the generated local review path shall at least include the review focus itself, and may selectively include its prerequisite knowledge points and / or subsequent knowledge points on the dependency chain, and the order of knowledge points in the path shall conform to the prerequisite logic of the knowledge graph. If it belongs to the relevance chain of the knowledge points to be reviewed, the generated local review path includes the review focus and one or more related knowledge points with high recommendation intensity; If they belong to two different relation chains, then a generation method is randomly determined.

8. The intention-driven personalized learning path recommendation method according to claim 6, characterized in that, The recommended strength is calculated using the following formula: ； in, K represents the knowledge points to be reviewed, and K represents the knowledge points currently being studied. These are the knowledge points to be reviewed. betweenness centrality, These are the knowledge points to be reviewed. The distance to the current learning knowledge point K, The user's attitude towards reviewing knowledge points is derived from the current knowledge status profile. The probability of mastering These are the knowledge points to be reviewed. Recommended strength.

9. A computer-readable storage medium, characterized in that, The storage medium stores a computer program that, when executed by a processor, implements the intent-driven personalized learning path recommendation method as described in any one of claims 1 to 8.

10. A computer electronic device, characterized in that, Including memory and processor; The memory is used to store computer programs; The processor is configured to, when executing the computer program, implement the intent-driven personalized learning path recommendation method as described in any one of claims 1 to 8.

Images