Search

CN-122019571-A - Learning resource recommendation method and device

CN122019571ACN 122019571 ACN122019571 ACN 122019571ACN-122019571-A

Abstract

The application discloses a learning resource recommendation method and device. The method comprises the steps of receiving a resource retrieval keyword which is input by a target object and comprises resource attributes of resources to be retrieved, obtaining a user image which comprises the grasping level of the target object to a plurality of knowledge points, constructing a Boolean retrieval expression according to the resource retrieval keyword and the user image, determining target learning resources matched with the Boolean retrieval expression according to a preset learning resource index, wherein the learning resource index comprises a Boolean logic expression which corresponds to the plurality of learning resources and comprises the multi-dimensional resource attributes corresponding to the learning resources, and the multi-dimensional resource attributes at least comprise the recommended grasping level of the learning resources. The application solves the technical problems that the traditional learning resource website lacks pertinence when recommending learning resources to users and cannot ensure the learning effect of the users.

Inventors

  • OUYANG WEI

Assignees

  • 湖南快乐阳光互动娱乐传媒有限公司

Dates

Publication Date
20260512
Application Date
20260130

Claims (10)

  1. 1. A learning resource recommendation method, comprising: receiving a resource search keyword input by a target object, and acquiring a user portrait of the target object, wherein the resource search keyword comprises resource attributes of resources to be searched, and the user portrait comprises mastering grades of the target object on a plurality of knowledge points, and each knowledge point corresponds to at least one learning resource; constructing a Boolean search expression according to the resource search keyword and the user portrait; Determining target learning resources matched with the Boolean search expressions according to preset learning resource indexes, wherein the learning resource indexes comprise Boolean logic expressions corresponding to a plurality of learning resources, the Boolean logic expressions comprise multi-dimensional resource attributes corresponding to the learning resources, and the multi-dimensional resource attributes at least comprise recommended mastering levels of the learning resources; recommending the target learning resource to the target object.
  2. 2. The method of claim 1, wherein the user representation is constructed in a manner comprising: obtaining learning records of the target object on a plurality of learning resources and answer records of supporting exercises of each learning resource, wherein each learning resource corresponds to one knowledge point and is provided with a set of exercises; For each knowledge point, determining first learning features of multiple dimensions corresponding to the knowledge point according to learning records and answering records corresponding to learning resources corresponding to the knowledge point, wherein the first learning features are used for reflecting the grasping degree of the target object on the knowledge point; carrying out weighted summation on the first learning features of the multiple dimensions according to preset weight coefficients to obtain first learning scores corresponding to the knowledge points; Determining a mastering grade corresponding to the first learning grade according to a preset grading grade mapping table, and taking the mastering grade as the mastering grade of the target object on the knowledge point, wherein the grading grade mapping table stores mapping relations between different first learning grading intervals and different mastering grades; And constructing the user portrait according to the grasping levels of the target object to a plurality of knowledge points.
  3. 3. The method of claim 2, wherein the learning resources are lesson videos, the learning records include video clip playback information, the answering records include answering time and accuracy, the dimensions of the first learning feature include at least one of an average answering accuracy, an answering stability index, learning efficiency, a knowledge grasping slope, and a learning burden index, and determining the first learning feature of the plurality of dimensions corresponding to the knowledge point according to the learning records and the answering records corresponding to the learning resources corresponding to the knowledge point includes: Determining a first average value of the correct rate in the answer records corresponding to all the learning resources corresponding to the knowledge points as the answer average correct rate, and/or, Determining standard deviation of correct rate in answer records corresponding to all learning resources corresponding to the knowledge points, determining a first ratio of the standard deviation to the first average value, determining a first difference of 1 and the first ratio as the answer stability index, and/or, Determining a second average value of answering time in answering records corresponding to all learning resources corresponding to the knowledge points, determining a second ratio of the second average value to preset standard answering time, determining a third ratio of the first average value to the second ratio as the learning efficiency, and/or, Respectively determining a third average value of the accuracy rate in the answer records corresponding to the learning resources with the highest recommended mastering level and a fourth average value of the accuracy rate in the answer records corresponding to the learning resources with the lowest recommended mastering level, determining a second difference value between the third average value and the fourth average value, determining a fourth ratio of the second difference value to the maximum span between the recommended mastering levels as the knowledge mastering slope, and/or, And determining a fifth ratio of the total duration of the video clips with the playback times exceeding a preset playback threshold value to the total duration of the complete video in the learning record corresponding to each learning resource corresponding to the knowledge point, and determining a fifth average value of a plurality of fifth ratios as the learning burden index.
  4. 4. The method of claim 1, wherein the determining of the recommended level of mastery of the learning resource comprises: For each learning resource, acquiring learning records of a plurality of objects on the learning resource and answer records of supporting problems of the learning resource, wherein each learning resource corresponds to one knowledge point and is provided with a set of problems; Determining second learning features of multiple dimensions corresponding to the learning resources according to the learning records and the answer records, wherein the second learning features are used for reflecting the mastering degree of different objects on knowledge points corresponding to the learning resources; weighting and summing the second learning characteristics of the multiple dimensions according to preset weight coefficients to obtain second learning scores corresponding to the knowledge points; Determining a recommended mastering level corresponding to the second learning score according to a preset scoring level mapping table, and taking the recommended mastering level as the recommended mastering level of the learning resource, wherein the scoring level mapping table stores mapping relations between different second learning scoring intervals and different recommended mastering levels.
  5. 5. The method of claim 4, wherein the learning resources are lesson videos, the learning records include video clip playback information, the answering records include answering time and accuracy, the dimensions of the second learning features include at least one of an average answering accuracy, a knowledge grasping slope, and a learning burden index, and determining the second learning features of the plurality of dimensions corresponding to the learning resources according to the learning records and the answering records includes: Determining weight coefficient of each object according to the grasping level of each object to the knowledge point corresponding to the learning resource, weighting and summing the correctness of each object in the answer records of the matched exercises of the learning resource according to the weight coefficient to obtain the average correctness of the answers, and/or, Determining a sixth average value of the accuracy rates in the answer records of the matched problems of the learning resources for each object with the highest mastering level of the knowledge points corresponding to the learning resources and a seventh average value of the accuracy rates in the answer records of the matched problems of the learning resources for each object with the lowest mastering level of the knowledge points corresponding to the learning resources, determining a third difference value between the sixth average value and the seventh average value, determining a sixth ratio of the third difference value to the maximum span between mastering levels as the knowledge mastering slope, and/or, And determining a seventh ratio of the total duration of the video clips, the number of which is larger than a preset playback threshold, in the learning record of the learning resource by each object to the total duration of the complete video, and determining an eighth average value of a plurality of seventh ratios as the learning burden index.
  6. 6. The method according to claim 1, wherein the learning resource index is constructed in a manner that includes: For each learning resource, acquiring a multi-dimensional resource attribute corresponding to the learning resource, constructing a sub-expression corresponding to the resource attribute of each dimension, and combining all the sub-expressions to obtain a Boolean logic expression corresponding to the learning resource, wherein the sub-expression is a triplet comprising a resource attribute name, a logic relationship and a resource attribute value, and the type of the logic relationship comprises belonging and non-belonging; Determining the conjunctive identifier and the conjunctive size of the Boolean logic expression corresponding to each learning resource, wherein the conjunctive identifiers corresponding to the same Boolean logic expression are the same, and the conjunctive size is the number of sub-expressions to which the corresponding logic relationship in the Boolean logic expression belongs; For the resource attribute of each dimension, forming a key binary group by a resource attribute name and a resource attribute value of the resource attribute, determining each Boolean logic expression containing a sub-expression corresponding to the resource attribute, forming a value binary group by a conjunctive identifier of each Boolean logic expression and a logic relationship in the sub-expression corresponding to the resource attribute in the Boolean logic expression, forming an inverted list by a plurality of value binary groups, and constructing a mapping relationship between the key binary group and the inverted list; constructing a first-level learning resource index according to the mapping relation between the key binary group corresponding to the resource attribute of each dimension and the inverted list; and constructing a secondary learning resource index according to the mapping relation between each joint identification and each corresponding learning resource.
  7. 7. The method of claim 6, wherein constructing a boolean search expression from the resource search key and the user representation comprises: Determining each resource attribute in the resource retrieval key word, and constructing a sub-retrieval formula corresponding to each resource attribute, wherein the resource attribute at least comprises a target knowledge point, and the sub-retrieval formula is a binary group comprising a resource attribute name and a resource attribute value; determining a target grasping level of the target object to the target knowledge point from the user portrait, and taking a binary group consisting of the target knowledge point and the target grasping level as a sub-search formula; And combining all the sub-search expressions to obtain the Boolean search expression.
  8. 8. The method of claim 7, wherein all sub-search expressions in the boolean search expression are conjunctive relationships, determining a target learning resource matching the boolean search expression according to a preset learning resource index, comprising: For each sub-search expression in the Boolean search expression, determining a key tuple matched with the sub-search expression from the first-level learning resource index, determining a reverse list corresponding to the key tuple, and sequencing all value tuples in the reverse list according to the joint size corresponding to the contained joint identification from small to large to obtain a value tuple set corresponding to the sub-search expression; Sorting the value binary group sets corresponding to the sub-search formulas in the Boolean search expression according to the joint size corresponding to the joint identification contained in the first value binary group in the set from small to large to obtain a value binary group set list; The following steps are circularly executed until the first value binary group set in the value binary group set list is empty: Starting from a second value binary group set in the value binary group set list, detecting whether a first value binary group in the current value binary group set is the same as a first value binary group in the previous value binary group set one by one; Determining a logical relationship in the same value tuple under the condition that the first value tuple in the set of all value tuples is detected to be the same; if the logic relation is not included, deleting a first value binary group in the binary group set of all values, and entering the next cycle; And if the logic relationship is that the logic relationship belongs to, determining that each learning resource corresponding to the conjunctive identifier in the same value binary group is the target learning resource from the secondary learning resource index, deleting the first value binary group in all value binary group sets, and entering the next cycle.
  9. 9. A learning resource recommendation device, comprising: The acquisition module is used for receiving a resource search keyword input by a target object and acquiring a user portrait of the target object, wherein the resource search keyword comprises resource attributes of resources to be searched, and the user portrait comprises mastering grades of the target object on a plurality of knowledge points, and each knowledge point corresponds to at least one learning resource; the construction module is used for constructing a Boolean search expression according to the resource search keywords and the user portrait; The matching module is used for determining target learning resources matched with the Boolean search expression according to a preset learning resource index, wherein the learning resource index comprises Boolean logic expressions corresponding to a plurality of learning resources, the Boolean logic expressions comprise multi-dimensional resource attributes corresponding to the learning resources, and the multi-dimensional resource attributes at least comprise recommended mastering grades of the learning resources; and the recommending module is used for recommending the target learning resource to the target object.
  10. 10. An electronic device comprising a memory and a processor, wherein the memory stores a computer program, and the processor is configured to execute the learning resource recommendation method according to any one of claims 1 to 8 by the computer program.

Description

Learning resource recommendation method and device Technical Field The application relates to the technical field of big data processing, in particular to a learning resource recommendation method and device. Background Under the background of rapid development of current digital education, an online learning platform has become an important channel for popularizing knowledge and improving skills. However, existing learning resource recommendation systems still have some drawbacks that limit the potential exertion of personalized education. Specifically, the conventional recommendation algorithm often depends on fixed content classification or trending, lacks deep insight into specific needs of individual learners, ignores the difference among learners, and fails to provide customized suggestions according to knowledge mastery degree, learning preference and cognitive ability of students. Especially in the field of education video courses, the mainstream platform usually ranks and recommends according to the praise, play amount or comment enthusiasm of the video, and neglects the matching degree between the actual educational value of course content and the current learning condition of students, and the recommendation mechanism may lead the students with lower performance to feel frustrated by the excessively complex materials, while the students with excellent performance lose interest due to content redundancy. More importantly, the lack of effective resource grading and matching algorithms makes it difficult for students to quickly find the most suitable learning path among a vast number of courses because a large number of high-quality educational resources cannot be effectively utilized. In addition, the prior art fails to achieve efficient retrieval and matching when dealing with large-scale user data and video resources, and the user experience is poor, especially when dealing with peak hours or facing global user groups, the probability of recommendation delay and mismatching is greatly increased. In view of the above problems, no effective solution has been proposed at present. Disclosure of Invention The embodiment of the application provides a learning resource recommending method and device, which at least solve the technical problems that the traditional learning resource website lacks pertinence when recommending learning resources to users and cannot guarantee the learning effect of the users. According to one aspect of the embodiment of the application, a learning resource recommendation method is provided, which comprises the steps of receiving a resource search keyword input by a target object, and obtaining a user portrait of the target object, wherein the resource search keyword comprises resource attributes of resources to be searched, the user portrait comprises mastery levels of the target object on a plurality of knowledge points, each knowledge point corresponds to at least one learning resource, building a Boolean search expression according to the resource search keyword and the user portrait, determining target learning resources matched with the Boolean search expression according to a preset learning resource index, wherein the learning resource index comprises Boolean logic expressions corresponding to the plurality of learning resources, the Boolean logic expressions comprise multi-dimensional resource attributes corresponding to the learning resources, and the multi-dimensional resource attributes at least comprise the recommended mastery levels of the learning resources. The user portrait construction method comprises the steps of obtaining learning records of a target object on a plurality of learning resources and answer records of supporting exercises of each learning resource, wherein each learning resource corresponds to one knowledge point and is provided with a set of exercises, determining first learning features of a plurality of dimensions corresponding to the knowledge points according to the learning records and the answer records corresponding to the learning resources corresponding to the knowledge points for each knowledge point, wherein the first learning features are used for reflecting the grasping degree of the target object on the knowledge points, carrying out weighted summation on the first learning features of the plurality of dimensions according to a preset weight coefficient to obtain first learning scores corresponding to the knowledge points, determining grasping grades corresponding to the first learning scores according to a preset grading mapping table, taking the grasping grades as grasping grades of the knowledge points of the target object, storing mapping relations between different first learning grading intervals and different grasping grades in the grading mapping table, and constructing the user portrait according to the grasping grades of the target object on the knowledge points. Optionally, the learning resources are course videos,