A method and system for recommending educational content based on swarm intelligence

Abstract

This invention discloses a method and system for recommending educational content based on swarm intelligence. The method includes: constructing a dynamic cognitive knowledge graph for users, where the memory strength attribute of knowledge point nodes is dynamically decayed based on a cognitive forgetting model and combined with memory stability; defining pheromones on the edges of the knowledge graph and strategically injecting exploration pheromones according to the user's forgetting state or knowledge breadth index; guiding a swarm intelligence algorithm to construct learning paths based on rules jointly influenced by bimodal pheromones; and finally determining the optimal path and generating recommendations. This invention, through a deep integration of cognitive science models and a bimodal pheromone-based swarm intelligence algorithm, achieves precise intervention in the learning process and intelligent guidance for knowledge exploration, significantly improving the stability of learning, effectively breaking down information cocoons, and enhancing the interpretability of recommendations through path intent analysis.

Description

Technical Field

[0001] This invention relates to the field of content recommendation technology, and in particular to a method and system for recommending educational content based on swarm intelligence. Background Technology

[0002] With the rapid development of information technology and the popularization of online education, educational resources have shifted from scarcity to overload. How to accurately and efficiently deliver personalized learning materials to learners from massive amounts of digital content has become a core challenge in the field of smart education. Early educational recommendation systems largely borrowed collaborative filtering and content-based recommendation algorithms from the e-commerce field, matching users' historical behavior or textual features of content to achieve a certain degree of initial personalization. However, due to the unique nature of educational scenarios—namely, the rigorous logical connections and cognitive dependencies inherent in knowledge itself—simple behavioral or content similarity matching is insufficient to meet the needs of structured learning. Therefore, the introduction of knowledge graphs has become a major technological breakthrough in this field. By constructing knowledge points and their preceding and succeeding relationships into a graph structure, recommendation systems can plan learning sequences for users along knowledge dependency paths, significantly improving the logic and systematic nature of recommendations. Building on this, to more precisely characterize learners' mastery status, researchers have introduced cognitive diagnostic models and knowledge tracing models. By analyzing users' interactive data such as answer records, they dynamically assess the probability of mastering each knowledge point, thereby achieving more precise individualized instruction.

[0003] Despite significant advancements in knowledge representation, user modeling, and path optimization, existing technologies still exhibit considerable limitations in simulating real-world human cognitive processes and guiding long-term learning. Firstly, current user cognitive models are relatively singular and static in their characterization, generally lacking a deep integration of the forgetting curve of human memory. Most methods focus solely on the user's "mastery" of knowledge points—whether they have learned them—while neglecting "memory stability," i.e., how long they can retain the information. According to the Ebbinghaus forgetting curve in cognitive science, the strength of knowledge retention decays exponentially over time, and repeated, timely review can significantly slow the rate of forgetting. However, existing recommendation methods cannot accurately model this dynamic decay process, resulting in recommended review content that is often arbitrary or based on fixed periods, rather than providing the most efficient intervention when the user's memory strength drops to a critical threshold. This leads to a lack of stability in the learner's knowledge system construction, making them prone to the phenomenon of "learning quickly, forgetting quickly." Secondly, existing recommendation algorithms generally suffer from the "exploitation-exploration dilemma," heavily favoring "exploitation." This easily leads learners into "information cocoons," and tends to delve deeply into and recommend related information within areas already familiar to the user. While this approach may improve learning efficiency and user satisfaction in the short term, it severely limits the breadth of learners' knowledge in the long run, hindering their ability to build cross-disciplinary knowledge connections and negatively impacting the development of innovative thinking and comprehensive abilities. Finally, most recommendation systems are like "black boxes," their recommendation logic opaque to the user. When a recommendation system presents a learning path to a user, it cannot explain the underlying "intention"—is this recommendation to reinforce old knowledge, overcome difficulties, or broaden horizons? This lack of explainability reduces user trust and learning initiative, making it difficult for learners to form a true "human-machine collaborative" learning partnership with the system. Summary of the Invention

[0004] The purpose of this section is to outline some aspects of embodiments of the present invention and to briefly describe some preferred embodiments. Simplifications or omissions may be made in this section, as well as in the abstract and title of this application, to avoid obscuring the purpose of these documents; however, such simplifications or omissions should not be construed as limiting the scope of the invention.

[0005] In view of the aforementioned existing problems, this invention is proposed. Therefore, this invention provides an educational content recommendation method based on swarm intelligence to solve the problems mentioned in the background section.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] In a first aspect, the present invention provides an educational content recommendation method based on swarm intelligence, comprising:

[0008] A dynamic cognitive knowledge graph is constructed for users, and the knowledge point nodes in the knowledge graph are associated with a memory strength attribute that decays over time based on a cognitive forgetting model;

[0009] Pheromones are defined on the edges of the knowledge graph, and exploratory pheromones are injected according to the user's dynamic cognitive state.

[0010] A group of agents in a swarm intelligence algorithm are guided to construct candidate learning paths on the knowledge graph based on a probability rule that is simultaneously influenced by the utilization of pheromones and the exploration of pheromones.

[0011] Based on the candidate learning paths, an optimal learning path is determined, and educational content recommendations corresponding to the knowledge point nodes on that path are generated.

[0012] As a preferred embodiment of the swarm intelligence-based educational content recommendation method of the present invention, the step of constructing a dynamic cognitive knowledge graph for the user and associating the knowledge point nodes in the knowledge graph with a memory strength attribute that decays over time based on a cognitive forgetting model further includes:

[0013] A memory stability parameter is dynamically updated based on the user's learning progress.

[0014] The decay rate of the memory strength attribute over time is adjusted using the memory stability parameter.

[0015] As a preferred embodiment of the swarm intelligence-based educational content recommendation method of the present invention, exploratory pheromones are injected when at least one of the following conditions is met:

[0016] Condition 1: Determine that the memory strength attribute of a certain knowledge point node has fallen below the first preset threshold;

[0017] Condition 2: Determine that the user's knowledge breadth index has fallen below the second preset threshold.

[0018] As a preferred embodiment of the swarm intelligence-based educational content recommendation method of the present invention, wherein: when condition 2 is satisfied, the injection of exploratory pheromones further includes:

[0019] The user's knowledge entropy is calculated as the knowledge breadth indicator;

[0020] Calculate the vector similarity between the centroids of knowledge points that the user has not mastered and those that the user has mastered, and determine the target path for injecting exploratory pheromones.

[0021] As a preferred embodiment of the swarm intelligence-based educational content recommendation method of the present invention, the construction of candidate learning paths includes:

[0022] Calculate a probability value that simultaneously includes the use of pheromones, exploration of pheromones, and heuristic information to select a candidate learning path;

[0023] The weights of pheromone utilization, pheromone exploration, and heuristic information in the calculation process are dynamically adjusted based on the user's current learning mode.

[0024] As a preferred embodiment of the swarm intelligence-based educational content recommendation method described in this invention, an asymmetric pheromone update process is further included after constructing candidate learning paths.

[0025] The agent performs enhancement and evaporation operations on the pheromone utilization based on the evaluation results of the candidate learning paths.

[0026] Once a knowledge point node has been learned by a user, the exploration pheromone pointing to that node is decayed or cleared.

[0027] As a preferred embodiment of the swarm intelligence-based educational content recommendation method of the present invention, the method further includes: determining an optimal learning path based on the candidate learning paths and generating educational content recommendations corresponding to the knowledge point nodes on that path;

[0028] Calculate the contribution of pheromone exploration on the optimal learning path to obtain the path intent score;

[0029] Based on the path intent score, an explanatory reason for this recommendation is generated by matching from a preset explanatory template library.

[0030] Secondly, the present invention provides an educational content recommendation system based on swarm intelligence, comprising:

[0031] The dynamic knowledge graph construction module is configured to build a dynamic cognitive knowledge graph for the user, and associate the knowledge point nodes in the knowledge graph with a memory strength attribute that decays over time based on a cognitive forgetting model;

[0032] The pheromone definition and injection module is configured to define utilization pheromones on the edges of the knowledge graph and inject exploration pheromones according to the user's dynamic cognitive state.

[0033] The learning path construction module is configured to guide a group of agents in a swarm intelligence algorithm to construct candidate learning paths on the knowledge graph based on a probability rule that is simultaneously influenced by the utilization pheromone and the exploration pheromone.

[0034] The recommendation module is configured to determine an optimal learning path based on the candidate learning paths and generate educational content recommendations corresponding to the knowledge point nodes on that path.

[0035] Thirdly, the present invention provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement any step of the above-described method.

[0036] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements any step of the above-described method.

[0037] Compared with existing technologies, the beneficial effects of the invention are:

[0038] 1. This invention introduces the Ebbinghaus forgetting curve and combines it with dynamically updated memory stability parameters to achieve accurate tracking of the user's memory state. This enables the recommendation system to provide review recommendations at key points of knowledge forgetting, thereby consolidating learning outcomes in the most efficient way and significantly enhancing the persistence of knowledge memory.

[0039] 2. By defining the use of pheromones and the exploration of pheromones, unlike the simple and random exploration strategies in existing technologies, this invention strategically injects exploration pheromones, which can balance the in-depth mining and breadth exploration of knowledge, effectively avoid the limitation of vision caused by over-exploitation, and thus purposefully guide users to explore new fields, promoting the integration of interdisciplinary knowledge and the cultivation of innovation capabilities.

[0040] 3. By quantifying the exploration intent and matching the explanation template, the teaching purpose behind each recommendation can be clearly explained to the user. This transparent design can transform the "black box" recommendation into an understandable human-computer dialogue, thereby enhancing the user's trust and learning initiative. Attached Figure Description

[0041] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Wherein:

[0042] Figure 1 This is a flowchart illustrating the overall process of a swarm intelligence-based educational content recommendation method according to an embodiment of the present invention. Detailed Implementation

[0043] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.

[0044] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0045] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0046] This invention is described in detail with reference to the schematic diagrams. When detailing the embodiments of this invention, for ease of explanation, the cross-sectional views illustrating the device structure may be partially enlarged, not adhering to the usual scale. Furthermore, the schematic diagrams are merely examples and should not be construed as limiting the scope of protection of this invention. In actual fabrication, the three-dimensional spatial dimensions of length, width, and depth should be included.

[0047] Furthermore, in the description of this invention, it should be noted that the terms "upper," "lower," "inner," and "outer," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. These terms are used solely for the convenience of describing the invention and for simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on the invention. In addition, the terms "first," "second," or "third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0048] Unless otherwise explicitly specified and limited, the terms "installation," "connection," and "joining" in this invention should be interpreted broadly. For example, they can refer to fixed connections, detachable connections, or integral connections; similarly, they can refer to mechanical connections, electrical connections, or direct connections, or indirect connections through an intermediate medium, or internal connections between two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0049] Example 1

[0050] Reference Figure 1This is the first embodiment of the present invention, which provides a method for recommending educational content based on swarm intelligence, including:

[0051] S1. Construct a dynamic cognitive knowledge graph for users, and associate the knowledge point nodes in the knowledge graph with a memory strength attribute that decays over time based on a cognitive forgetting model;

[0052] Specifically, construct a general knowledge graph for one or more subject areas. This general knowledge graph is in the form of a directed graph: G = (V, E), where the set of nodes V represents all knowledge points in the subject area, such as "Newton's second law" and "Pythagorean theorem", and the set of edges E represents the dependencies between knowledge points, that is, the preconditions, successors or related relationships. For example, the edge from "quadratic equation" to "quadratic function" indicates that the former is the learning basis of the latter.

[0053] It should be noted that this map can be manually constructed by experts in the field of education, or semi-automatically generated from materials such as textbooks and syllabi using text mining algorithms;

[0054] Furthermore, a user state layer is added to each user on top of this general knowledge graph to construct a dynamic cognitive knowledge graph for the user;

[0055] Specifically, through this user state layer, each knowledge point node v∈V in the dynamic cognitive knowledge graph is assigned a dynamic attribute set, where each attribute in the attribute set has a definite value at each time point t:

[0056] Mastery: A value in the range [0,1], calculated by combining user behavior data such as test scores, practice accuracy, and learning time related to this knowledge point;

[0057] Recent access timestamp: The time when a user last had a valid interaction with content related to this knowledge point;

[0058] Memory Intensity: Based on a cognitive forgetting model, this invention reflects the persistence of knowledge points in memory. Specifically, the preferred cognitive forgetting model is the Ebbinghaus forgetting curve. However, unlike the Ebbinghaus forgetting curve, this invention defines memory intensity as a function that dynamically changes with time t, with a value range between [0,1], where 1 represents the deepest memory and 0 represents complete forgetting. The decay law of this function strictly follows the exponential form of the Ebbinghaus forgetting curve. When the user at time t... last After completing a study or review of a certain knowledge point, the strength of memory at any future time t (t>t) last The value of ) can be calculated using the following formula:

[0059] M(t) = exp(-k×(tt)last ))

[0060] Where M(t) is the memory strength function, exp is the natural exponential function, and tt last It is represented as the time interval since the last learning, and k is a constant characterizing the forgetting rate;

[0061] It should be noted that the initial value of memory strength is 0 or a very small value, indicating that the user has not yet learned it;

[0062] Furthermore, in order to more accurately simulate the complexity of user learning, the present invention also dynamically updates a memory stability based on the user's learning effect in the traditional Ebbinghaus forgetting curve, and uses the memory stability parameter to adjust the decay rate of memory strength attribute over time.

[0063] Specifically, memory stability is defined as a positive number to quantify how "firmly" a user has grasped a knowledge point. The higher the memory stability, the less likely the knowledge point is to be forgotten. The update of memory stability is mainly based on the user's learning performance.

[0064] Whenever a user successfully completes a learning session or quiz (e.g., answers a question correctly in a quiz), the memory stability of that knowledge point is increased accordingly. When a user fails to learn (e.g., answers a question incorrectly), the memory stability may remain unchanged or decrease slightly.

[0065] Preferably, the increase in memory stability is positively correlated with learning effectiveness. For example, the increase in memory stability resulting from answering a question correctly on the first attempt is greater than the increase resulting from answering a question correctly after multiple attempts.

[0066] Specifically, by adjusting the decay rate using memory stability, the forgetting rate k in the memory strength function is transformed into a function of memory stability. Let k = 1 / S, where S represents memory stability. Then, the calculation formula for the memory strength function M(t) is modified as follows:

[0067] M(t) = exp(-(tt) last ) / S)

[0068] It should be noted that by introducing memory stability to adjust memory intensity, this invention achieves scientific dynamic cognitive knowledge graph modeling. That is, for a newly learned knowledge point, its initial memory stability is low, so 1 / S is large, and the memory intensity function decays very quickly, which is in line with the general rule that "new knowledge is easy to forget". However, as the user reviews the knowledge point multiple times and successfully, its memory stability will continue to increase, resulting in 1 / S becoming smaller. The decay curve of the memory intensity function will become flatter and flatter, accurately simulating the cognitive science principle that review can "combat" forgetting and extend the memory period.

[0069] S2. Define the use of pheromones on the edges of the knowledge graph and inject exploration pheromones according to the user's dynamic cognitive state;

[0070] It should be noted that, taking into account the depth and breadth of users' knowledge learning, this invention defines the use of pheromones and the exploration of pheromones based on the constructed dynamic cognitive knowledge graph;

[0071] Furthermore, pheromones will be defined on each directed edge (i, j) of the knowledge graph, and the magnitude of the pheromone will represent the recommendation strength for migration from knowledge point i to knowledge point j.

[0072] It should be noted that the purpose of utilizing pheromones is to guide the agent to deeply explore and consolidate the user's existing knowledge network, that is, to execute the "utilization" strategy. Furthermore, the update of pheromones is an adaptive update, that is, to enhance or decay (evaporate) them according to the quality of the learning paths generated during the algorithm iteration process. This is a typical positive feedback mechanism, which makes the concentration of pheromones on high-quality learning paths higher and higher, so that they are selected by subsequent agents with a higher probability.

[0073] Furthermore, unlike the pheromone used in algorithmic iterative updates, the defined exploratory pheromone is a guiding signal injected into the learning path by a user's cognitive state. The role of this exploratory pheromone is to guide the agent to explore knowledge areas that are ignored under the current "utilization" strategy but are helpful for the user's long-term learning, that is, to execute the "exploration" strategy, thereby effectively breaking the information cocoon.

[0074] Furthermore, in order to achieve an intelligent "exploration" strategy, the injection of exploration pheromones will be triggered when at least one of the following preset conditions is met:

[0075] Condition 1: Determine that the memory strength attribute (i.e., memory strength function value) of a certain knowledge point node has fallen below the first preset threshold;

[0076] It should be noted that this condition aims to achieve accurate review based on the Ebbinghaus forgetting curve. By periodically or before each recommendation, the user's dynamic cognitive knowledge graph is traversed to check the current memory strength of all learned knowledge points (i.e., nodes with M>0). If the memory strength of a certain knowledge point is found to be less than a preset threshold (e.g., 0.4), it is determined that the knowledge point has entered the "forgetting critical zone". At this time, in order to guide the algorithm to prioritize planning a review path that includes the knowledge point, positive exploration pheromones are injected on all edges pointing to the node. This is equivalent to setting a temporary "highlight beacon" for the knowledge point to be reviewed on the knowledge graph to ensure that it receives higher attention in the next round of learning path construction.

[0077] Condition 2: Determine that the user's knowledge breadth index has fallen below the second preset threshold;

[0078] It should be noted that this condition is mainly used to promote the expansion of knowledge breadth in order to assess the breadth or diversity of the user's current knowledge system. When it is found that the user's learning is too focused on a narrow field and may be trapped in an information cocoon, the user will be guided to explore new fields.

[0079] Preferably, in order to achieve accurate judgment and effective execution of condition 2, the user's knowledge entropy is calculated as an indicator of knowledge breadth.

[0080] It should be explained that knowledge entropy is an indicator that measures the uncertainty of a system. In this invention, it is mainly used to quantify the degree of balance in the distribution of user knowledge. Assuming that the knowledge points already mastered by the user are distributed across N different sub-disciplines (such as mechanics, electricity, and optics in physics), p n Let H be the proportion of the number of knowledge points a user has mastered in the nth subdomain to the total number of knowledge points they have mastered. Then, the user's knowledge entropy H can be calculated and expressed as:

[0081] H=-Σ(p n ×log(p n ))

[0082] The formula represents summing over all N subdomains; the higher the knowledge entropy value, the more balanced and broad the distribution of user knowledge; the lower the knowledge entropy value, the more concentrated and poorer the distribution of knowledge.

[0083] Furthermore, the calculated knowledge entropy is compared with a preset threshold. If the knowledge entropy is less than the preset threshold, the exploration mechanism is triggered. Once the exploration mechanism is triggered, the target path for injecting exploration pheromones is determined by calculating the vector similarity between the centroids of the knowledge points that the user has not mastered and the knowledge domains that the user has mastered, so as to ensure the relevance and acceptability of the "exploration" rather than a completely random jump.

[0084] Specifically, firstly, pre-trained language models such as Word2Vec and BERT are used to map all knowledge points (including those already mastered and those not yet mastered) in the knowledge graph to the same high-dimensional vector space, where each knowledge point is represented by a vector, and the distance and angle between vectors can reflect the semantic relevance between knowledge points. For subdomains where user knowledge distribution is too concentrated, the average of all mastered knowledge point vectors in that domain is calculated to obtain the knowledge centroid vector of that domain. Then, all unmastered knowledge points are traversed, and the cosine similarity between the vector of each unmastered knowledge point and the knowledge centroid vector is calculated. By selecting knowledge points that are neither completely overlapping with nor completely unrelated to the user's current knowledge core area as optimal exploration targets, for example, selecting unmastered knowledge points with similarity within an intermediate range, exploration pheromones are injected into the path pointed to by the user's current knowledge boundary if the optimal exploration target knowledge point is determined.

[0085] It should be noted that by defining the utilization and exploration of pheromones, not only can the user's knowledge depth be consolidated, but also the user can be guided to conduct efficient knowledge exploration when forgetting or having too narrow a knowledge base, thereby fundamentally solving the "utilization-exploration" dilemma in existing technologies;

[0086] S3. Guide a group of agents in the swarm intelligence algorithm to construct candidate learning paths on the knowledge graph based on a probability rule that is simultaneously influenced by the utilization of pheromones and the exploration of pheromones.

[0087] Furthermore, by simulating swarm intelligence behavior, multiple high-quality candidate learning paths can be efficiently searched on a knowledge graph that integrates user cognitive state and guidance intentions.

[0088] Preferably, the swarm intelligence algorithm is the Ant Colony Optimization (ACO) algorithm or a variant thereof, where the intelligent agent is the corresponding ant;

[0089] Specifically, a group of agents (i.e., ant colony) is initialized, generating M agents, and these M agents are placed on the knowledge graph. The starting position of the agent is the knowledge point that the user is currently learning, or a node on the knowledge boundary determined according to the user's learning history. The agents construct paths according to probability rules. Each agent at a node in the knowledge graph needs to decide the next knowledge point node to visit, and this decision-making process is not random but determined by probability rules.

[0090] Furthermore, a probability value that simultaneously incorporates pheromone utilization, pheromone exploration, and heuristic information is calculated to select candidate learning paths;

[0091] Specifically, this probability value is obtained from a probability rule, which represents the probability that an agent will move from node i to its adjacent, unvisited node j, and is derived from the following formula:

[0092]

[0093] Among them, P m (i,j) represents the transition probability of the m-th agent (ant) choosing knowledge point node j as the next knowledge point node to visit when it is currently at knowledge point node i; τ(i,j) represents the concentration of utilization pheromone defined on the directed edge (i,j), reflecting the historical success experience and recommendation strength of the path from node i to node j; φ(i,j) represents the exploration pheromone concentration defined on the directed edge, obtained by actively injecting guidance signals based on the user's dynamic cognitive state (such as forgetting or narrow knowledge); ω τ and ω φ These represent the dynamic weights for utilizing pheromones and exploring pheromones, respectively. These two weights are dynamically adjusted based on the user's current learning mode (e.g., "reinforcement and review mode," "new knowledge exploration mode," or "regular learning mode"), satisfying ω τ +ω φ =1, enabling the algorithm to flexibly switch its strategic focus between "utilization" and "exploration"; η(i,j) represents the heuristic information for the transition from node i to node j, also known as "visibility," which is represented as a priori expectation that does not change with pheromones. In this invention, η(i,j) can be flexibly defined according to the needs of the educational scenario. For example, it can be defined as the reciprocal of the learning value of the target knowledge point j, such as η(i,j) = 1 / D j D j It is the difficulty coefficient of knowledge point j, or preferably, it is related to the cognitive state of the target knowledge point, for example, defined as η(i,j)=1-M j (t), meaning the lower the memory strength of the target knowledge point j, the higher its heuristic value; α and β are two preset positive constants, used as exponents of the pheromone factor and the heuristic factor, respectively; where α controls the relative importance of the pheromone trajectory, the larger its value, the more the agent tends to choose the path with high pheromone concentration, and the stronger the convergence of the algorithm; while β controls the relative importance of heuristic information, the larger the β value, the more the agent tends to choose the node with high heuristic information value, and the exploratory and random nature of the algorithm will be enhanced; This represents the set of all allowed neighboring nodes that agent m can access when it is at node i. Nodes that agent m has already visited in this round of path construction (recorded in the "taboo list") are excluded to avoid path loops.

[0094] Furthermore, based on the user's current learning pattern, the weights of utilizing pheromones, exploring pheromones, and heuristic information in the calculation process are dynamically adjusted.

[0095] Specifically, in the consolidation and review mode, when it is detected that the user's memory strength for multiple knowledge points has decreased, it is determined that the user needs to consolidate the knowledge points. At this time, the weight of exploring pheromones is increased.

[0096] Specifically, in the new knowledge exploration mode, when a user's knowledge entropy decreases, it is necessary to broaden their horizons and increase the weight of exploration pheromones and heuristic information.

[0097] Specifically, in the normal learning mode, when the user's cognitive state is stable, the weight of pheromone utilization is increased, while the weight of heuristic information is reduced.

[0098] Furthermore, after the candidate learning path is constructed, each agent starts from the starting point and uses a "tacit list" to record the visited nodes to prevent path looping. It continuously selects the next node according to the above probability rules until three termination conditions are met (i.e., the path length reaches a preset value, the path covers a certain number of unmastered knowledge points, or the total learning time budget is exhausted).

[0099] Specifically, once all agents have completed their respective path construction, a set containing different candidate learning paths can be obtained, which is the same number as the total number of agents.

[0100] S4. Determine an optimal learning path based on the candidate learning paths, and generate educational content recommendations corresponding to the knowledge point nodes on that path.

[0101] Furthermore, all candidate learning paths are evaluated by creating a fitness function, as shown in the following formula:

[0102] Fitness (Path) m )=ω gain ·V gain (Path m )+ω review ·V review (Path m )-ω cost ·C cost (Path m )

[0103] Among them, Fitness (Path) m ) represents the candidate learning path constructed by the m-th agent (ant). m The fitness score indicates that the path has greater overall value to the current user, and is therefore more likely to be selected as the optimal learning path; Vgain (Path m V represents the knowledge gain value of a candidate learning path, used to measure the path's contribution to guiding users to learn new knowledge; it is the sum of all knowledge points on the candidate learning path that the user has not yet mastered. review (Path m The value of a candidate learning path is represented by (C), which measures its contribution to helping users consolidate learned knowledge and combat forgetting. It is the sum of the forgetting levels of all learned knowledge points along the candidate learning path (i.e., the lower the memory strength of the knowledge points traversed by a path, the closer they are to being forgotten, and the higher their review value). cost (Path m ω represents the learning cost of a candidate learning path, used to quantify the time and effort required for that path. It is the sum of the estimated learning times for all knowledge points on the candidate learning path, or the sum of the difficulty coefficients of all knowledge points. gain ω review and ω cost These are the weighting coefficients for knowledge gain, review value, and learning cost, respectively, satisfying ω gain +ω review +ω cost =1;

[0104] Furthermore, among all candidate learning paths, the candidate learning path with the highest fitness score is selected as the optimal solution for this iteration;

[0105] In addition, after determining the optimal path, the pheromones on the knowledge graph need to be updated to prepare for the next recommendation. The following two methods can be used for updating:

[0106] The first approach involves performing enhancement and evaporation operations using pheromones. This operation follows the classic ant colony algorithm update rules: a fixed evaporation rate ρ (a constant between 0 and 1) is defined for all edges in the knowledge graph to decay the pheromone content. The decay process is represented as follows:

[0107] τ(i,j)←(1-ρ)×τ(i,j)

[0108] It should be noted that this operation is used to prevent the algorithm from converging to a local optimum too early, and to give the algorithm the possibility of continuously exploring new paths;

[0109] Subsequently, pheromone enhancement is applied to the edges traversed by the best-performing paths in this iteration, with the pheromone increment proportional to the path's fitness.

[0110] Δτ(i,j)=Q / max(Fitness(Path m ))

[0111] Where Δτ(i,j) is the pheromone increment, and Q is a positive constant used to control the total amount of pheromone increment;

[0112] Combining the decay process and pheromone increment described above, the new pheromone of the edge on the learning path with the highest fitness is represented as:

[0113] τ(i,j)←τ(i,j)+Δτ(i,j)

[0114] The second type is triggered when a user learns a knowledge point node, that is, when the user completes the learning interaction of a certain knowledge point in the learning path with the highest adaptability (such as watching a video or doing the exercises correctly), the exploration pheromone is updated.

[0115] Since the exploration pheromone is essentially a temporary guiding beacon, once its guiding task is completed (i.e., the user has learned the knowledge point according to the guidance), the beacon should be removed to avoid permanent and unnecessary interference with subsequent path planning. Therefore, assuming the user has just learned the knowledge point on the learning path with the highest fitness, it is only necessary to find all the edges pointing to that node and directly attenuate or clear the exploration pheromone on these edges to zero.

[0116] Furthermore, the contribution of pheromones explored along the optimal learning path is calculated to obtain the path intent score, which is the ratio of the contribution of pheromones explored along all edges of the optimal learning path to the total contribution of pheromones.

[0117]

[0118] Wherein, IS(max(Fitness(Path) m ))) represents the path intent score;

[0119] Furthermore, a pre-built explanation template library is provided, which contains humanized explanation statements for different recommendation intentions;

[0120] Specifically, when the path intent score is high (e.g., >0.7) and the exploration pheromone was injected because of condition 1, the matching precise review template will provide an explanatory statement, such as "We noticed that your memory of knowledge point A may be a little fuzzy. This learning path will help you consolidate your knowledge efficiently while learning new knowledge."

[0121] Specifically, when the path intent score is high (e.g., >0.7) and the exploration pheromone was injected because of condition 2, the broadened horizons template is matched to give an explanatory statement, such as "You are already very solid in the field of mechanics! In order to broaden your knowledge horizons, we recommend that you explore the related field of energy conservation. This learning path can help you build a more comprehensive view of physics."

[0122] Specifically, when the path intent score is very low (e.g., <0.3), a deep mining template is matched to provide an explanation statement, such as "Based on your recent learning progress, this learning path will guide you to explore quadratic functions more deeply and overcome the key points and difficulties."

[0123] Furthermore, this embodiment also provides an educational content recommendation system based on swarm intelligence, including:

[0124] The dynamic knowledge graph construction module is configured to build a dynamic cognitive knowledge graph for the user, and associate the knowledge point nodes in the knowledge graph with a memory strength attribute that decays over time based on a cognitive forgetting model;

[0125] The pheromone definition and injection module is configured to define utilization pheromones on the edges of the knowledge graph and inject exploration pheromones according to the user's dynamic cognitive state.

[0126] The learning path construction module is configured to guide a group of agents in a swarm intelligence algorithm to construct candidate learning paths on the knowledge graph based on a probability rule that is simultaneously influenced by the utilization pheromone and the exploration pheromone.

[0127] The recommendation module is configured to determine an optimal learning path based on the candidate learning paths and generate educational content recommendations corresponding to the knowledge point nodes on that path.

[0128] This embodiment also provides a computer device suitable for educational content recommendation methods based on swarm intelligence, including:

[0129] The system includes a memory and a processor. The memory stores computer-executable instructions, and the processor executes these instructions to implement the swarm intelligence-based educational content recommendation method proposed in the above embodiments.

[0130] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0131] This embodiment also provides a storage medium storing a computer program that, when executed by a processor, implements the swarm intelligence-based educational content recommendation method proposed in the above embodiments.

[0132] The storage medium proposed in this embodiment and the data storage method proposed in the above embodiments belong to the same inventive concept. Technical details not described in detail in this embodiment can be found in the above embodiments, and this embodiment has the same beneficial effects as the above embodiments.

[0133] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code. The solutions in the embodiments of this application can be implemented using various computer languages, such as the object-oriented programming language Java and the interpreted scripting language JavaScript.

[0134] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0135] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0136] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0137] Although preferred embodiments of this application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of this application.

[0138] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. A method for recommending educational content based on swarm intelligence, characterized in that, include: A dynamic cognitive knowledge graph is constructed for users, and the knowledge point nodes in the knowledge graph are associated with a memory strength attribute that decays over time based on a cognitive forgetting model; Pheromones are defined on the edges of the knowledge graph, and exploratory pheromones are injected according to the user's dynamic cognitive state. Pheromones are defined on each directed edge of the knowledge graph. The magnitude of the pheromone represents the recommendation strength for migrating from one knowledge point to another. The pheromone is used to guide the agent to conduct in-depth mining and consolidation along the user's existing knowledge network. The exploration pheromone is a guiding signal injected into the learning path by a user's cognitive state. The exploration pheromone is used to guide the agent to explore knowledge areas that are ignored under the current utilization strategy but are helpful for the user's long-term learning. Exploratory pheromones are injected when at least one of the following conditions is met: Condition 1: Determine that the memory strength attribute of a certain knowledge point node has fallen below the first preset threshold; Condition 2: Determine that the user's knowledge breadth index has fallen below the second preset threshold; A group of agents in a swarm intelligence algorithm are guided to construct candidate learning paths on the knowledge graph based on a probability rule that is simultaneously influenced by the utilization of pheromones and the exploration of pheromones. The construction of candidate learning paths includes: Calculate a probability value that simultaneously includes the use of pheromones, exploration of pheromones, and heuristic information to select a candidate learning path; The probability value is obtained from a probability rule, which represents the probability that the agent will move from node i to its adjacent, unvisited node j, and is derived from the following formula: in, This represents the transition probability of the m-th agent choosing the next node to visit as j when it is currently located at node i. This represents the concentration of pheromones used on a directed edge; This represents the pheromone concentration for exploration defined on a directed edge; and These are represented as the dynamic weights for utilizing pheromones and exploring pheromones, respectively. This represents the heuristic information indicating the transition from node i to node j; and Two pre-defined positive constants are used as indices for the pheromone factor and the heuristic factor, respectively. Let m represent the set of all neighboring nodes that agent m is allowed to access at node i. The weights of utilizing pheromones, exploring pheromones, and heuristic information in the calculation process are dynamically adjusted based on the user's current learning mode. Based on the candidate learning paths, an optimal learning path is determined, and educational content recommendations corresponding to the knowledge point nodes on that path are generated.

2. The educational content recommendation method based on swarm intelligence as described in claim 1, characterized in that, The method of constructing a dynamic cognitive knowledge graph for users, and associating knowledge point nodes in the knowledge graph with a memory strength attribute that decays over time based on a cognitive forgetting model, also includes: A memory stability parameter is dynamically updated based on the user's learning progress. The decay rate of the memory strength attribute over time is adjusted using the memory stability parameter.

3. The educational content recommendation method based on swarm intelligence as described in claim 1, characterized in that, When condition 2 is met, the injection of exploratory pheromones also includes: The user's knowledge entropy is calculated as the knowledge breadth indicator; Calculate the vector similarity between the centroids of knowledge points that the user has not mastered and those that the user has mastered, and determine the target path for injecting exploratory pheromones.

4. The educational content recommendation method based on swarm intelligence as described in claim 1, characterized in that, Following the construction of candidate learning paths, an asymmetric pheromone update process is also included: The agent performs enhancement and evaporation operations on the pheromone utilization based on the evaluation results of the candidate learning paths. Once a knowledge point node has been learned by a user, the exploration pheromone pointing to that node is decayed or cleared.

5. The educational content recommendation method based on swarm intelligence as described in claim 1, characterized in that, Determining an optimal learning path based on the candidate learning paths and generating educational content recommendations corresponding to the knowledge point nodes on that path also includes: Calculate the contribution of pheromone exploration on the optimal learning path to obtain the path intent score; Based on the path intent score, an explanatory reason for this recommendation is generated by matching from a preset explanatory template library.

6. A swarm intelligence-based educational content recommendation system, based on the swarm intelligence-based educational content recommendation method according to any one of claims 1 to 5, characterized in that, include: The dynamic knowledge graph construction module is configured to build a dynamic cognitive knowledge graph for the user, and associate the knowledge point nodes in the knowledge graph with a memory strength attribute that decays over time based on a cognitive forgetting model; The pheromone definition and injection module is configured to define utilization pheromones on the edges of the knowledge graph and inject exploration pheromones according to the user's dynamic cognitive state. Pheromones are defined on each directed edge of the knowledge graph. The magnitude of the pheromone represents the recommendation strength for migrating from one knowledge point to another. The pheromone is used to guide the agent to conduct in-depth mining and consolidation along the user's existing knowledge network. The exploration pheromone is a guiding signal injected into the learning path by a user's cognitive state. The exploration pheromone is used to guide the agent to explore knowledge areas that are ignored under the current utilization strategy but are helpful for the user's long-term learning. Exploratory pheromones are injected when at least one of the following conditions is met: Condition 1: Determine that the memory strength attribute of a certain knowledge point node has fallen below the first preset threshold; Condition 2: Determine that the user's knowledge breadth index has fallen below the second preset threshold; The learning path construction module is configured to guide a group of agents in a swarm intelligence algorithm to construct candidate learning paths on the knowledge graph based on a probability rule that is simultaneously influenced by the utilization pheromone and the exploration pheromone. The construction of candidate learning paths includes: Calculate a probability value that simultaneously includes the use of pheromones, exploration of pheromones, and heuristic information to select a candidate learning path; The probability value is obtained from a probability rule, which represents the probability that the agent will move from node i to its adjacent, unvisited node j, and is derived from the following formula: in, This represents the transition probability of the m-th agent choosing the next node to visit as j when it is currently located at node i. This represents the concentration of pheromones used on a directed edge; This represents the pheromone concentration for exploration defined on a directed edge; and These are represented as the dynamic weights for utilizing pheromones and exploring pheromones, respectively. This represents the heuristic information indicating the transition from node i to node j; and Two pre-defined positive constants are used as indices for the pheromone factor and the heuristic factor, respectively. Let m represent the set of all neighboring nodes that agent m is allowed to access at node i. The weights of utilizing pheromones, exploring pheromones, and heuristic information in the calculation process are dynamically adjusted based on the user's current learning mode. The recommendation module is configured to determine an optimal learning path based on the candidate learning paths and generate educational content recommendations corresponding to the knowledge point nodes on that path.

7. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 5.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 5.

Images