An online interaction method and system

Abstract

The application belongs to the technical field of data processing, and particularly relates to an online interaction method and system, which comprises the following steps: constructing a recommended learning path according to a course knowledge graph by using an ant colony algorithm, obtaining learning target inspiration, value inspiration, difficulty inspiration and interest inspiration of a course in each round of iteration of the ant colony algorithm, and then constructing an inspiration function of the course; obtaining pheromone of edges between all courses according to the number of courses contained in each path in the last round of iteration of the ant colony algorithm and the standard learning duration of the courses; determining selection probabilities of optional courses according to the pheromone of edges between the course corresponding to the current node and each optional course and the inspiration function of each optional course, and selecting an optional course as a next node of the path according to the selection probabilities. The application realizes personalized learning path recommendation, and can improve the learning efficiency and knowledge mastery effect of students.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing. More particularly, the present application relates to an online interaction method and system. BACKGROUND

[0002] With the continuous advancement of educational informatization, personalized learning path recommendation technology has become an important means to improve teaching efficiency and meet the differentiated needs of students. Currently, some researches have attempted to combine knowledge modeling and recommendation algorithms to provide customized learning paths for students. However, most of these methods still have obvious limitations.

[0003] Most existing methods rely on simple rule matching or linear ranking strategies to generate learning paths, lack in-depth modeling of complex semantic relationships between courses and knowledge points, and are difficult to accurately reflect the overall structure of the knowledge system. This leads to the recommended learning paths often ignoring prerequisite knowledge requirements, causing cognitive gaps in the learning process, and affecting the depth of understanding and learning coherence of students.

[0004] In addition, existing methods often have weak depiction of individual differences among students, usually only based on interest tags for rough recommendations, lack comprehensive analysis of multi-dimensional information such as students' learning behavior characteristics and mastery of knowledge points, and are difficult to achieve true "teaching students in accordance with their aptitude". At the same time, some recommendation mechanisms do not have dynamic adjustment capabilities and cannot optimize the path in real time according to the students' learning progress, reducing the adaptability and practicality of the recommended results.

[0005] Therefore, it is urgent to introduce a more intelligent and efficient path generation method that can deeply integrate knowledge structure and student characteristics and has dynamic response capabilities to significantly improve the accuracy and practicality of personalized learning path recommendation. SUMMARY

[0006] To solve the technical problems that the existing methods are difficult to accurately reflect the overall structure of the knowledge system, difficult to achieve teaching students in accordance with their aptitude, and do not have dynamic adjustment capabilities, the present application provides solutions in the following aspects.

[0007] In a first aspect, the present application provides an online interaction method, comprising:

[0008] The course knowledge graph is constructed; the recommended learning path is constructed by using an ant colony algorithm according to the course knowledge graph; the recommended learning path is visualized in the course knowledge graph, and students are guided to learn in sequence; wherein in each round of iteration process of the ant colony algorithm: the learning target heuristic is obtained according to the importance of the knowledge points contained in the course and the relevance of the learning target and the field to which the knowledge points belong; the value heuristic is obtained according to the mastery degree of the students to the learned knowledge points and the relevance of the learned knowledge points and the knowledge points contained in the course; the difficulty heuristic is obtained according to the difficulty of each knowledge point contained in the course; the interest heuristic is obtained according to the preference index of the students to each resource type; the heuristic function of the course is constructed according to the learning target heuristic, the value heuristic, the difficulty heuristic and the interest heuristic; the pheromone of the edges between all courses is obtained according to the number of courses contained in each path in the last iteration result of the ant colony algorithm and the standard learning time length of the courses; the courses that meet the precondition, have not been learned by the students and have not been added to the path are selected as optional courses, the selection probability of each optional course is determined according to the pheromone of the edges between the course corresponding to the current node and each optional course and the heuristic function of each optional course, and the optional course is selected as the next node of the path according to the selection probability.

[0009] The present application realizes semantic modeling of courses, knowledge points and their mutual relations by constructing a structured course knowledge graph, makes the knowledge system more clear and calculable. On this basis, combined with multi-dimensional information such as learning behavior, learning achievement, mastery degree and interest preference of students, the learning target and weak link of students can be accurately identified, and comprehensive data support is provided for personalized path recommendation. The present application uses an ant colony algorithm to plan a learning path, fully considers factors such as logical coherence of the path, learning efficiency, difficulty adaptability and student interest, can dynamically generate a high-quality, personalized learning path under the premise of meeting the prerequisite dependency relationship, has good self-adaptability and search ability, improves the accuracy and practicability of the recommended result, and solves the problems such as path breakage and cognitive discontinuity in traditional methods. The present application dynamically adjusts the recommended path according to the learning progress of students, enhances the flexibility and interactivity of the recommended mechanism, and improves the intelligent level of the learning path recommendation.

[0010] Preferably, the learning target heuristic is obtained by: performing text preprocessing on the learning target of the student, and converting the preprocessed text into a plurality of word vectors; converting the field to which each knowledge point contained in the course belongs into a word vector; calculating the average of the cosine similarity between each word vector corresponding to the learning target and the word vector corresponding to the field to which the knowledge point belongs, as the relevance of the learning target and the field to which the knowledge point belongs; and performing weighted summation on the relevance of the learning target and the field to which the knowledge point belongs according to the importance of the knowledge points contained in the course, to obtain the learning target heuristic of the course.

[0011] This method effectively measures the degree of match between learning objectives and course content in semantic space by vectorizing the semantic information of learning objectives and course knowledge points and calculating the cosine similarity between the two. Furthermore, the correlation results are weighted and summed, taking into account the importance of the knowledge points. This allows the learning objectives of the course to be more closely aligned with the actual knowledge structure and student needs, thereby improving the accuracy and personalization of recommended paths and helping students prioritize courses that are highly aligned with their learning goals.

[0012] Preferably, the value heuristic satisfies the expression:

[0013] ;in, Indicates the value inspiration of course K; Indicates the first knowledge points; Indicates the learned knowledge points; Indicates the learned Knowledge points and the first The relevance of each knowledge point; Indicates the number of knowledge points included in course K; Indicates the number of knowledge points learned; Indicates the students' understanding of the The degree of mastery of each knowledge point.

[0014] This method effectively measures the value of a course to a student's knowledge system by comprehensively considering the correlation between already learned knowledge points and the knowledge points in the selected courses, as well as the student's mastery of the already learned knowledge points. Knowledge points with lower mastery levels receive higher learning demand weights, while courses containing knowledge points with higher correlations receive higher value inspiration. This encourages students to prioritize content closely related to their weak points, achieving the goal of strengthening understanding through repeated learning and gradually addressing knowledge gaps, thereby improving the targeted and personalized level of path recommendations.

[0015] Preferably, the difficulty heuristic satisfies the expression:

[0016] ;in, Indicates the difficulty inspiration of course K; Indicates the learned Knowledge points and the first The relevance of each knowledge point; Indicates the number of knowledge points included in course K; Indicates the number of knowledge points learned; Indicates the students' understanding of the The degree of mastery of each knowledge point; Indicates the first The difficulty of each knowledge point; is the maximum value function.

[0017] This method dynamically assesses the actual learning difficulty of a course for students by comprehensively considering the inherent difficulty of each knowledge point in the course and the student's mastery of the relevant knowledge points. For each knowledge point, the most relevant already mastered knowledge point is selected and weighted according to the mastery level. This makes the difficulty assessment more consistent with the student's actual cognitive level. It can reasonably avoid courses that are too difficult for students and lack sufficient pre-requisite support in the recommended path, thereby improving the adaptability and acceptability of the learning path and helping students complete knowledge construction step by step.

[0018] Preferably, the obtaining of interest inspiration includes: taking the resource type of the course as the target resource type, and taking the student's preference index for the target resource type as the interest inspiration of the course.

[0019] Preferably, the heuristic function for constructing a course includes weighted summing of the normalized results of learning goal inspiration, the normalized results of value inspiration, the normalized results of interest inspiration, and the negative correlation normalized results of difficulty inspiration to obtain the heuristic function of the course.

[0020] The present invention normalizes learning goal inspiration, value inspiration, interest inspiration and difficulty inspiration, and performs weighted fusion according to set weights to construct a comprehensive inspiration function, which can realize the organic integration of multi-dimensional information, taking into account the matching degree between courses and learning goals, the supplementary value of knowledge weaknesses, students' interest preferences and the learning difficulty of the course. The final inspiration function reflects the degree of adaptation of the course to the student's learning path more comprehensively and accurately, and provides a scientific and personalized decision-making basis for the recommendation system.

[0021] Preferably, the method for obtaining pheromones for edges between courses includes: obtaining a quality score of the path based on the number of courses included in the path and the standard learning duration of the course; for any two courses, if the two courses are adjacent nodes in a path of the previous iteration result, then the corresponding path is used as a reference path for the two courses; and based on the quality scores of all reference paths of the two courses, the pheromones for the edges between the two courses are updated: , Indicates the course With courses The pheromones between the edges; Indicates the course in the previous iteration With courses The pheromones between the edges; representing a course with a course an average of quality scores of all reference paths between courses; representing pheromone evaporation factor.

[0022] The application can effectively reflect the value of transition between different courses by comprehensively considering the number of courses in the path and the standard learning duration, calculating the quality score of the path, and dynamically adjusting the pheromone intensity of the edge based on the co-occurrence of adjacent courses in historical paths, thereby improving the recognition ability of efficient learning paths and enhancing the stability and adaptability of path recommendation, which helps to guide students to construct more reasonable and coherent learning sequences.

[0023] Preferably, the method for obtaining the degree of mastery of the student over the learned knowledge points comprises: for any one learned knowledge point, obtaining the average of simulation scores corresponding to all simulation operation tasks related to the knowledge point, normalizing the average of simulation scores to obtain the operation proficiency of the student over the knowledge point; taking the accuracy of the student over the knowledge point in the stage test and examination as the theoretical proficiency of the student over the knowledge point, and performing weighted average on the operation proficiency and the theoretical proficiency to obtain the degree of mastery of the student over the knowledge point.

[0024] Preferably, the method for obtaining the preference index of the student for each resource type comprises: for any one resource type, taking the learning duration, access frequency, task completion rate, collection frequency, viewing frequency and simulation operation frequency of the student on all courses of the resource type as one evaluation dimension of the resource type, respectively, performing normalization on each evaluation dimension, and performing weighted summation on the normalized results of each evaluation dimension to obtain the preference index of the student for the resource type.

[0025] In a second aspect, the application provides an online interaction system, comprising a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the above-mentioned online interaction method is realized.

[0026] By adopting the above technical solution, the above-mentioned online interaction method is generated into a computer program and stored in the memory to be loaded and executed by the processor, so that a terminal device is manufactured according to the memory and the processor, and use is facilitated.

[0027] The application has the following beneficial effects:

[0028] The present application realizes semantic modeling of courses, knowledge points and their mutual relations by constructing a structured course knowledge graph, so that the knowledge system is more clear and computable. On this basis, combined with multi-dimensional information such as students' learning behavior, learning performance, mastery degree and interest preference, the learning target and weak link of the students can be accurately identified, and comprehensive data support is provided for personalized path recommendation. The present application adopts ant colony algorithm for learning path planning, fully considers factors such as logical coherence of the path, learning efficiency, difficulty adaptability and student interest, can dynamically generate high-quality and personalized learning paths under the premise of meeting the pre-dependence relationship, has good self-adaptability and search ability, improves the accuracy and practicability of the recommended results, and solves the problems such as path breakage and cognitive fault in the traditional method. The present application dynamically adjusts the recommended path according to the learning progress of the students, enhances the flexibility and interactivity of the recommended mechanism, and improves the intelligent level of the learning path recommendation. BRIEF DESCRIPTION OF DRAWINGS

[0029] Figure 1 is a flowchart schematically showing an online interaction method in the present application;

[0030] Figure 2 is a flowchart schematically showing path construction of each round of ant colony algorithm. DETAILED DESCRIPTION

[0031] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0032] The specific embodiments of the present application will be described in detail below with reference to the drawings.

[0033] The embodiments of the present application disclose an online interaction method, referring to Figure 1 , comprising steps S1-S4:

[0034] S1, constructing a course knowledge graph.

[0035] Specifically, data of various sources such as teaching outline, teaching materials, teaching objectives, teaching contents and past papers are collected, and the original data are subjected to text preprocessing, including word segmentation, part-of-speech tagging, text cleaning and entity standardization and the like. Based on natural language processing technology, core entities such as “course” and “knowledge point” are identified, and the semantic relationship (such as preposition relationship, inclusion relationship, correlation, etc.) between entities and the attribute information of each entity are further extracted.

[0036] The attributes of a course include: the field to which it belongs, the type of resources (such as graphics, text, video, audio, simulation), standard learning duration, prerequisite courses, and included knowledge points; the attributes of knowledge points include: difficulty, field to which it belongs, course to which it belongs, importance, and relevance to other knowledge points.

[0037] Among them, the difficulty of knowledge points is quantified based on the average accuracy rate of knowledge points in previous years' examination papers. The lower the accuracy rate, the higher the difficulty, and the value range is [0,1]. The importance of knowledge points is evaluated by the frequency of their appearance in the teaching syllabus and examination syllabus. The higher the frequency of appearance, the higher the importance, and the value range is [0,1]. The correlation between knowledge points is automatically extracted through co-occurrence analysis and semantic similarity model, and then manually corrected and supplemented by educational experts to ensure the accuracy of the logical relationship between knowledge points, and the value range is [0,1].

[0038] Finally, the extracted entities, relationships and their attributes are organized in a graph structure to construct a course knowledge graph.

[0039] S2. Collect students' learning behaviors, academic performance, and learning goals, obtain students' mastery of learned knowledge points based on their academic performance, and construct students' preference index for various resource types based on their learning behaviors.

[0040] Specifically, for any student, learning information in multiple dimensions such as learning behavior, academic performance, and learning goals is collected.

[0041] Learning behavior mainly includes students' learning time, access frequency, task completion rate and interactive behavior (such as collection, repeated viewing, and number of simulation operations) in each course.

[0042] Learning performance includes two categories: theoretical performance and simulation performance. The theoretical performance comes from the periodic tests and exams corresponding to courses whose resource types are graphics, text, video, or audio, and includes the students' accuracy rate of each knowledge point in the periodic tests and exams; the simulation performance comes from the simulation operation tasks set in courses whose resource types are simulation.

[0043] Learning objectives refer to the degree of knowledge mastery or skill improvement that students hope to achieve through learning, and are set by students themselves, such as "becoming familiar with the functions and usage of control cores such as PLCs and microcontrollers", "mastering the installation, debugging and data acquisition methods of various sensors", "understanding the basic structure and working principles of control systems", etc.

[0044] Further, by analyzing the theoretical scores and simulation scores of the students, the mastery degree of each knowledge point learned by the students is calculated comprehensively, which is used to measure the understanding and application level of the students on the knowledge point. By counting the learning time distribution, access frequency, task completion rate and interactive behavior (such as collection, repeated viewing, simulation operation times) of the students on the courses of different resource types (text, video, audio, simulation), the preference index of the students on each resource type is constructed, which is used to reflect the preference degree of the students on the courses of different resource types. For example, if a student actively completes the simulation task many times and stays for a long time, it can be judged that he / she is more inclined to practical learning; if he / she watches the video course frequently, it indicates that he / she prefers audio-visual resources.

[0045] In one embodiment, the method for obtaining the mastery degree of each knowledge point learned by the students includes: for any one of the learned knowledge points, obtaining the mean value of the simulation scores corresponding to all simulation operation tasks related to the knowledge point, normalizing the mean value to obtain the operation proficiency of the student on the knowledge point, taking the correct rate of the student on the knowledge point in the stage test and examination as the theoretical proficiency of the student on the knowledge point, and performing weighted average on the operation proficiency and the theoretical proficiency to obtain the mastery degree of the student on the knowledge point. Wherein, when performing weighted average on the operation proficiency and the theoretical proficiency, the weights of the operation proficiency and the theoretical proficiency are set by the implementer according to the nature of the course, for example, if the course emphasizes theoretical understanding, the weight of the operation proficiency can be set to 0.4 and the weight of the theoretical proficiency can be set to 0.6, if the course emphasizes practical ability, the weight of the operation proficiency can be set to 0.6 and the weight of the theoretical proficiency can be set to 0.4. It should be noted that in order to ensure the rationality of the weighted average, the sum of the weights of the operation proficiency and the theoretical proficiency should be 1.

[0046] It should be noted that when the student does not perform the simulation operation task related to the knowledge point, the simulation score of the student on the corresponding simulation operation task is set to 0; when the knowledge point has no related simulation operation task, the theoretical proficiency of the student on the knowledge point is taken as the mastery degree of the student on the knowledge point; when the student does not participate in the stage test and examination, or the content of the stage test and examination does not contain the knowledge point, the theoretical proficiency of the student on the knowledge point is set to 0; when the content of all stage tests and examinations of the course does not contain the knowledge point, the operation proficiency of the student on the knowledge point is taken as the mastery degree of the student on the knowledge point. When the knowledge point has no related simulation operation task and the content of all stage tests and examinations of the course does not contain the knowledge point, the mastery degree of the student on the knowledge point is set to 0.

[0047] In one embodiment, the method for constructing the preference index of the students on each resource type includes:

[0048] For any resource type, the learning duration, the access times, the task completion rate, the collection times, the watching times and the simulation operation times of the student on all courses of the resource type are respectively taken as a judgment dimension of the resource type, each judgment dimension is normalized, and the results of the normalization of each judgment dimension are weighted and summed to obtain the preference index of the student to the resource type. The normalization of each judgment dimension includes: for any judgment dimension, the data of the judgment dimension of the resource type is divided by the sum of the data of the judgment dimension of all resource types to realize the normalization of the data of the judgment dimension of the resource type.

[0049] It should be noted that when the results of the normalization of each judgment dimension are weighted and summed in this embodiment, the weights of each judgment dimension are the same, and are , represents the number of judgment dimensions, and in other embodiments, the implementer can set the weight of each judgment dimension according to the actual implementation situation.

[0050] S3, according to the learning goal of the student, the mastery degree of the student to the learned knowledge points, the preference index of the student to each resource type and the course knowledge graph, a recommended learning path is constructed by using an ant colony algorithm.

[0051] Specifically, in each iteration process of the ant colony algorithm, the last course learned by the student is taken as a starting node, and the construction of each round of path is started from the starting node. In response to that the iteration times of the ant colony algorithm reach the maximum iteration times or the optimal paths obtained by continuous H rounds of iteration are the same, the ant colony algorithm is stopped, the starting node in the finally obtained optimal path is removed, and the remaining path is taken as the recommended learning path. Wherein is a preset number, in an embodiment, is set to 5, the maximum iteration times in the ant colony algorithm is 100, the number of ants is 50, the pheromone factor is 1, the heuristic function factor is 1, the pheromone evaporation factor is 1, and the initial pheromone of the edge between any two courses is 1, and in other embodiments, the implementer can set them according to the actual implementation situation.

[0052] It should be noted that the last course learned by the student is taken as the starting node in this application, and the construction of each round of path is started from the starting node, which can guarantee the continuity and logicality of the learning path, ensure that the newly recommended course has good connection with the existing learning basis in the knowledge structure, and thus improve the rationality and effectiveness of the personalized recommendation. The starting node is removed in the finally obtained optimal path to avoid repeated recommendation of learned content, and the output learning path only contains course recommendation results that have not been learned but have a high degree of matching with the learning goal.

[0053] Further, the flowchart of each round of path construction refers to Figure 2 , including steps S301 to S308, specifically:

[0054] S301, obtaining a learning goal hint of the course according to importance of knowledge points contained in the course and relevance of the learning goal to the domain of the knowledge points contained in the course.

[0055] S302, obtaining a value hint of the course according to the mastery degree of the learned knowledge points and the relevance of the learned knowledge points to the knowledge points contained in the course.

[0056] S303, obtaining a difficulty hint of the course according to the mastery degree of the learned knowledge points, the relevance of the learned knowledge points to the knowledge points contained in the course, and the difficulty of each knowledge point in the course.

[0057] S304, obtaining an interest hint of the course according to the resource type of the course and the preference index of the student to each resource type.

[0058] S305, constructing a hint function of the course according to the learning goal hint, the value hint, the difficulty hint, and the interest hint of the course.

[0059] S306, obtaining pheromone of edges between all courses according to the number of courses contained in each path in the last iteration result of the ant colony algorithm and the standard learning duration of the courses.

[0060] S307, taking the course that meets the precondition, has not been learned by the student, and has not been added to the path as a selectable course, determining selection probability of each selectable course according to the pheromone of edges between the course corresponding to the current node and each selectable course and the hint function of each selectable course, and selecting the selectable course as the next node of the path according to the selection probability.

[0061] S308, repeating step S307 until the current iteration of the ant colony algorithm is stopped when the stop condition is met, and the construction of the current round of path is implemented.

[0062] In one embodiment, the learning goal hint of the course is obtained according to the importance of the knowledge points contained in the course and the relevance of the learning goal to the domain of the knowledge points contained in the course in step S301, including:

[0063] Perform text preprocessing on students' learning objectives, including word segmentation and stop word removal. Use a natural language processing (NLP) model to convert the preprocessed text into several word vectors. Use the NLP model to convert the domains of each knowledge point included in the course into word vectors. Calculate the mean cosine similarity between the word vectors corresponding to the learning objectives and the word vectors corresponding to the domains to which the knowledge points belong. Calculate the mean cosine similarity between the word vectors corresponding to the learning objectives and the domains to which the knowledge points belong. Determine the learning objectives of the course based on the importance of the knowledge points included in the course and the relevance of the learning objectives to the domains to which the knowledge points belong:

[0064] ;

[0065] in, It represents the learning goal inspiration of course K; Indicates the first The importance of each knowledge point; Indicates the number of knowledge points included in course K; Indicates the student's learning goals, Indicates the first knowledge points; Indicates the student's learning objectives and the The correlation between the learning objectives and the fields to which the knowledge points belong is obtained based on the mean of cosine similarity, and the value range of cosine similarity is [-1,1]. The value range of is also [-1,1], so this embodiment adopts Correlation Perform normalization. Indicates the first The importance weight of each knowledge point. When the student's learning goal is more relevant to the field of the more important knowledge point in course K, the learning goal of course K is more inspiring, and course K should be the student's priority course.

[0066] It should be noted that the NLP model used to obtain word vectors in this embodiment is the Word2Vec model. In other embodiments, implementers can select an NLP model according to actual implementation conditions, such as BERT.

[0067] In one embodiment, step S302 obtains the value inspiration of the course based on the student's mastery of the learned knowledge points and the relevance of the learned knowledge points to the knowledge points included in the course, including:

[0068] The value heuristic satisfies the expression:

[0069] ;

[0070] wherein, represents the value enlightenment of the course K; represents the th knowledge point in the course K; represents the th learned knowledge point; represents the relevance of the th learned knowledge point and the th knowledge point in the course K; represents the number of knowledge points contained in the course K; represents the number of learned knowledge points; represents the mastery degree of the th learned knowledge point by the student; when the mastery degree of the th learned knowledge point is smaller, more attention is paid to the course to which the knowledge point related thereto belongs, thereby encouraging the student to learn the course related to the weak knowledge point and improving the mastery degree of the knowledge point through repeated learning of the related knowledge point.

[0071] In one embodiment, the difficulty enlightenment of the course is obtained according to the mastery degree of the learned knowledge point by the student, the relevance of the learned knowledge point and the knowledge point contained in the course, and the difficulty of each knowledge point in the course in step S303, comprising:

[0072] The difficulty enlightenment satisfies the expression:

[0073] ;

[0074] wherein, represents the difficulty enlightenment of the course K; represents the relevance of the th learned knowledge point and the th knowledge point in the course K; represents the number of knowledge points contained in the course K; represents the number of learned knowledge points; represents the mastery degree of the th learned knowledge point by the student; represents the difficulty of the th knowledge point in the course K; is a maximum function; when the relevance of the learned knowledge point and the th knowledge point in the course K is greater, and the mastery degree of the learned knowledge point by the student is greater, the difficulty of the th knowledge point in the course K is relatively reduced for the student, and therefore the th knowledge point in the course K is taken as the The difficulty reduction coefficient of each knowledge point is used to obtain the first The actual learning difficulty of each knowledge point , the actual learning difficulty of all knowledge points in Course K is comprehensively considered to obtain inspiration for the difficulty of Course K.

[0075] In one embodiment, step S304 obtains interest inspiration for the course based on the resource types of the course and the student's preference index for each resource type, including:

[0076] The resource type of the course is used as the target resource type, and the students' preference index for the target resource type is used as the interest inspiration for the course.

[0077] In one embodiment, step S305 constructs a course heuristic function based on the learning goal heuristic, value heuristic, difficulty heuristic, and interest heuristic of the course, including:

[0078] The heuristic function of the course satisfies the expression:

[0079] ;

[0080] in, represents the heuristic function of course K; 、 、 、 They represent the weight of learning goal inspiration, value inspiration, difficulty inspiration, and interest inspiration respectively; Indicates the learning objectives of course K The normalized result of Indicates the value inspiration of course K The normalized result of Indicates the difficulty of course K The reciprocal of The normalized result of Expressed interest in course K In this embodiment, the learning objectives of all courses that students have not yet learned are used to inspire the learning objectives of course K. Perform maximum and minimum value normalization and use the value inspiration of all courses that students have not yet learned to inspire the value of course K Perform maximum and minimum value normalization and use the inverse of the difficulty of all courses that students have not yet learned to inspire the difficulty of course K The reciprocal of is normalized to the maximum and minimum values, and the interest of all courses that students have not yet learned is used to inspire their interest in course K. The maximum-minimum value normalization is performed, and in other embodiments, the implementer can select other normalization methods according to actual implementation conditions, for example, maximum value normalization; when the learning goal heuristic of the course K is greater, the value heuristic is greater, the difficulty heuristic is smaller, and the interest heuristic is greater, the heuristic function of the course K is greater, encouraging the student to preferentially learn the course that is highly matched with the learning goal, helps to fill the weak point knowledge, has smaller difficulty, and meets the personal interest preference.

[0081] It should be noted that, since the learning goal heuristic has a core guiding role in path recommendation, directly affecting the matching degree of the course and the student's target knowledge system, the weight of the learning goal heuristic in this embodiment is set to 0.4, the weight of the value heuristic is set to 0.2, the weight of the difficulty heuristic is set to 0.2, and the weight of the interest heuristic is set to 0.2. In other embodiments, the implementer can set the weights of the learning goal heuristic, the value heuristic, the difficulty heuristic, and the interest heuristic according to actual implementation conditions.

[0082] In one embodiment, in step S306, the pheromone of the edge between all courses is obtained according to the number of courses contained in each path in the last iteration result of the ant colony algorithm and the standard learning duration of the courses, including:

[0083] For any path in the last iteration result, the quality score of the path is obtained according to the number of courses contained in the path and the standard learning duration of the courses:

[0084]

[0085] wherein, Q represents the quality score of the path; N represents the number of courses contained in the path; T represents the sum of the standard learning durations of all courses contained in the path, that is, the total standard learning duration of the path; exp represents an exponential function with a natural constant as the base, which is used for negative correlation mapping of N; and norm represents a normalization function. ​​​​​​​​​​​​​​​​​​​Perform maximum and minimum value normalization. In other embodiments, the implementer may select other normalization methods according to the actual implementation situation, such as maximum value normalization. The fewer courses a path contains, the The shorter the total standard learning time, the better the path The higher the learning efficiency, the better the path The higher the quality score.

[0086] For any two courses, if they are adjacent nodes in a path in the previous iteration, the corresponding path will be used as the reference path for the two courses. Based on the quality scores of all reference paths of the two courses, the pheromone of the edge between the two courses is updated:

[0087] ;

[0088] in, Indicates the course With courses The pheromones between the edges; Indicates the course in the previous iteration With courses The pheromones between the edges; Indicates the course With courses The mean of the quality scores of all reference paths between the courses With courses When there is no reference path between is 0; Indicates pheromone volatility factor.

[0089] In one embodiment, satisfying the prerequisite in step S307 means that all the prerequisite courses of the course have been learned or added to the path.

[0090] In one embodiment, in step S307, the selection probability of each optional course is determined based on the pheromone of the edge between the course corresponding to the current node and each optional course and the heuristic function of each optional course, including:

[0091] ;

[0092] Among them, the course corresponding to the current node is used To express, Indicates that optional courses will be The probability of selecting the next node as the current node; Indicates the course and optional courses The pheromones between the edges; Indicates optional courses a heuristic function; denotes a set of optional courses; denotes a pheromone factor; denotes a heuristic function factor. When a course has a larger pheromone between the optional course and the optional course has a larger heuristic function, the selection probability of the optional course , thereby encouraging students to preferentially learn courses that are highly matched with their learning goals, help to make up for weak point knowledge, have smaller difficulty, conform to personal interest preferences, and have higher overall learning efficiency.

[0093] In one embodiment, the step S307 of selecting an optional course as the next node of the path according to the selection probability comprises:

[0094] In a manner of sampling with unequal probabilities, an optional course is selected as the next node of the path from all optional courses according to the selection probability of each optional course.

[0095] In one embodiment, the setting method of the stop condition in the step S308 is:

[0096] The method in the step S301 is used to obtain the relevance of the learning goal of the student to the domain of all knowledge points, and all knowledge points with a relevance greater than a preset relevance threshold are taken as knowledge points covered by the learning goal. In response to the knowledge points covered by the courses included in the path already including all knowledge points covered by the learning goal, the construction of the path is stopped, and in response to all paths corresponding to each ant having been stopped from being constructed, the current iteration is ended.

[0097] The relevance threshold is set by an implementer according to actual implementation conditions, and in the embodiment, the relevance threshold is 0.3.

[0098] S4, visualizing the recommended learning path in the knowledge graph and guiding the student to learn in sequence.

[0099] Specifically, the recommended learning path is explicitly identified in the knowledge graph and displayed to the student in a visualized form, helping the student to intuitively understand the logical relationship between the courses and the knowledge points and the learning order between the courses, and suggesting the student to learn the courses according to the recommended learning path, so as to ensure the step-by-step and systematic mastery of knowledge and thereby improve the learning efficiency and effect.

[0100] The embodiment of the application further discloses an online interaction system comprising a processor and a memory, and the memory stores computer program instructions which, when executed by the processor, implement the online interaction method according to the application.

[0101] The system also comprises other components well known to those skilled in the art such as a communication bus and a communication interface, the arrangement and function of which are known in the art and thus will not be described here in more detail.

Claims

1. An online interaction method, characterized in that, The method comprises the following steps: Collecting data, identifying core entities based on natural language processing technology, further extracting semantic relationships between entities and attribute relationships of each core entity; the core entities include courses and knowledge points; organizing the extracted entities, semantic relationships and attributes in the form of a graph structure to construct a course knowledge graph; According to the course knowledge graph, a recommended learning path is constructed using an ant colony algorithm; the recommended learning path is visualized in the course knowledge graph to guide students to learn in sequence; In each iteration process of the ant colony algorithm: According to the importance of the knowledge points included in the course and the relevance of the student's learning goal and the domain of the knowledge points, the learning goal heuristic is obtained; according to the mastery degree of the student on the learned knowledge points and the relevance of the learned knowledge points and the knowledge points included in the course, the value heuristic is obtained; according to the difficulty of each knowledge point included in the course, the difficulty heuristic is obtained; according to the preference index of the student on each resource type, the interest heuristic is obtained; the heuristic function of the course is constructed according to the learning goal heuristic, the value heuristic, the difficulty heuristic and the interest heuristic; According to the number of courses included in each path in the last iteration result of the ant colony algorithm, the standard learning time of the course, the pheromone of the edge between all courses is obtained; the courses that meet the precondition, the student has not learned and have not been added to the path are selected as optional courses; according to the pheromone of the edge between the course corresponding to the current node and each optional course and the heuristic function of each optional course, the selection probability of each optional course is determined, and the optional course is selected as the next node of the path according to the selection probability; the method for obtaining the pheromone of the edge between the courses comprises: According to the number of courses contained in the path and the standard learning duration of the courses, a quality score of the path is obtained; for any two courses, if the two courses are adjacent nodes in a path in the result of the last iteration, the path is taken as a reference path of the two courses; according to the quality scores of all reference paths between the two courses, the pheromone of the edge between the two courses is updated: , denotes the pheromone of the edge between course and course ; denotes the pheromone of the edge between course and course in the last iteration; denotes the average of the quality scores of all reference paths between course and course ; denotes the pheromone evaporation factor.

2. The method of claim 1, wherein, The method for obtaining the learning goal heuristic comprises: Text preprocessing is performed on the learning goal of the student, and the preprocessed text is converted into a plurality of word vectors; the domain of each knowledge point included in the course is converted into a word vector; the average value of the cosine similarity between each word vector corresponding to the learning goal and the word vector corresponding to the domain of the knowledge point is calculated as the relevance of the learning goal and the domain of the knowledge point; the relevance of the learning goal and the domain of the knowledge point is weighted and summed according to the importance of the knowledge points included in the course to obtain the learning goal heuristic of the course.

3. The method of claim 1, wherein, The value heuristic satisfies the expression: ； wherein, represents the value enlightenment of the course K; represents the th knowledge point in the course K; represents the th learned knowledge point; represents the th learned knowledge point and the th knowledge point in the course K; represents the number of knowledge points contained in the course K; represents the number of learned knowledge points; represents the mastery degree of the th learned knowledge point by the student.

4. The method of claim 1, wherein, The difficulty heuristic satisfies the expression: ； wherein, represents the difficulty heuristic of course K; represents the relevance of the i-th learned knowledge point to the i-th knowledge point in course K; represents the relevance of the i-th learned knowledge point to the i-th knowledge point in course K; represents the number of knowledge points contained in course K; represents the number of learned knowledge points; represents the number of learned knowledge points; represents the mastery level of the i-th learned knowledge point by the student; represents the mastery level of the i-th learned knowledge point by the student; represents the difficulty of the i-th knowledge point in course K; represents the difficulty of the i-th knowledge point in course K; is a maximum function.

5. The method of claim 1, wherein, The method for obtaining the interest heuristic comprises: The resource type of the course is taken as the target resource type, and the preference index of the student on the target resource type is taken as the interest heuristic of the course.

6. The method of claim 1, wherein, The method for constructing the heuristic function of the course comprises: The normalized result of the learning goal heuristic, the normalized result of the value heuristic, the normalized result of the interest heuristic and the negatively correlated normalized result of the difficulty heuristic are weighted and summed to obtain the heuristic function of the course.

7. The method of claim 1 or 3 or 4, wherein, The method for obtaining the mastery degree of the student on the learned knowledge points comprises: For any one learned knowledge point, the mean value of simulation scores corresponding to all simulation operation tasks related to the knowledge point of the student is obtained, the mean value of the simulation scores is normalized to obtain the operation proficiency of the student on the knowledge point; the correctness rate of the student on the knowledge point in the stage test and examination is taken as the theoretical proficiency of the student on the knowledge point, and the operation proficiency and the theoretical proficiency are weighted and averaged to obtain the mastery degree of the student on the knowledge point.

8. The method of claim 1 or 5, wherein, The method for obtaining the preference index of the student for each resource type comprises: For any one resource type, the learning time, access times, task completion rate, collection times, viewing times and simulation operation times of the student on all courses of the resource type are respectively taken as one evaluation dimension of the resource type, each evaluation dimension is normalized, and the weighted sum of the normalized results of each evaluation dimension is obtained to obtain the preference index of the student for the resource type.

9. An online interactive system, characterized by It comprises: A processor and a memory, the memory stores computer program instructions, when the computer program instructions are executed by the processor, a kind of online interaction method according to any one of claims 1-8 is realized.

Images