Learning resource recommendation method and system based on personalized recommendation
By constructing user profiles and using text analysis technology, the relevance between learning resources and user interest characteristics is assessed, solving the problem of the difficulty in recommending learning resources in a targeted manner in existing technologies. This enables personalized learning resource recommendations, improving learning efficiency and user satisfaction.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-23
- Publication Date
- 2026-03-27
AI Technical Summary
Existing technologies struggle to recommend learning resources tailored to a user's learning and work fields, and also fail to provide relevant recommendations based on various dimensions of learning resources. This makes it difficult for users to quickly find high-quality learning resources that meet their needs.
By constructing user profiles and extracting feature information from learning resources, and using text analysis and natural language processing techniques, the correlation between learning resources and user interest features is evaluated, and learning resources that meet user needs are selected.
It enables personalized learning resource recommendations, improves learning efficiency and satisfaction, ensures that the resources users obtain are more in line with their interests and learning needs, and saves search time and effort.
Smart Images
Figure CN121743592A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of personalized recommendation, specifically a learning resource recommendation method and system based on personalized recommendation. Background Technology
[0002] With the rapid development of internet technology and the arrival of the information age, online learning resources are becoming increasingly abundant. Users often face information overload when choosing suitable learning resources. The sheer volume of resources makes it difficult for users to quickly find high-quality content that meets their needs, thus affecting learning efficiency and experience. As the online education market continues to expand and users' demands for personalized learning increase, more and more educational institutions and companies are beginning to focus on the research and application of learning resource recommendation systems. Users hope that the system can recommend the most suitable learning resources based on their learning goals and interests to improve learning efficiency and effectiveness.
[0003] Most recommendations are based solely on user browsing history and are not tailored to individual interests. They fail to consider the user's specific learning or work field to provide more targeted learning resources. Furthermore, they struggle to recommend more relevant resources based on various dimensions of the learning resources. Summary of the Invention
[0004] The present invention aims to at least solve one of the technical problems existing in the prior art; to this end, the present invention proposes a learning resource recommendation method and system based on personalized recommendation, which is used to solve the technical problems that it is difficult to take into account the user's own learning field or work field to make more targeted recommendations for learning resources, and at the same time, it is difficult to make more relevant resource recommendations based on various dimensions of learning resources.
[0005] To address the above problems, a first aspect of the present invention provides a learning resource recommendation method and system based on personalized recommendation, comprising the following steps: Collect users' learning information to construct user profiles, and extract learning feature domains based on user profiles and user browsing history; at the same time, extract feature information from learning resource data, and extract knowledge domain information from feature information; The data in the learning resource database is filtered to select learning resources whose knowledge domains match the user's learning characteristics. Extract resource feature data from learning resource data, extract user learning interest feature data from user browsing history, and evaluate the correlation between resource feature data and user learning interest feature data in a single screening of learning resource data through text analysis and natural language processing techniques. Based on the relevance evaluation results, the learning resource data selected in the first screening is further filtered, and a recommended list of learning materials is constructed by selecting the learning resources.
[0006] As a further aspect of the present invention: user learning information is collected to construct user profile information, and learning feature domains are extracted based on the user profile information and user browsing history; simultaneously, feature information of learning resource data is extracted, and knowledge domain information from the feature information is extracted, including the following steps: Collect learning information submitted by users, including: education level, occupation, professional field and interest field, to build user profile information; By analyzing users' historical learning behavior, we can obtain their browsing history, collection history, and evaluation history. The browsing history includes: viewing history, click history, and viewing duration. By analyzing the fields of education and occupation information in user profiles, we extract the fields associated with user profile information; we also extract the knowledge fields of learning materials corresponding to user browsing history, collection history, and evaluation history as the fields associated with browsing history. The user's professional field, interest field, user profile information association field, and browsing history association field are used as the user's learning feature fields; Collect learning resource data, establish a learning resource database, perform structured processing on the learning resource data, and extract the characteristic information of the learning resource data, including: title, abstract, text content, and knowledge domain.
[0007] As a further aspect of the present invention: extracting resource feature data from learning resource data, and extracting user learning interest feature data through user browsing history, includes the following steps: Obtain feature information from learning resource data; Using the TextRank algorithm, we extract the title, summary, text content and keyword information in the knowledge domain from the feature information, and use the keyword information as the resource feature data of the learning resource data. We obtain feature information of corresponding learning materials from users' browsing history, collection history, and evaluation history. We then use the TextRank algorithm to extract keyword information of titles, abstracts, text content, and knowledge domains from the feature information of the corresponding learning materials, which serves as the user's learning interest feature data.
[0008] As a further aspect of the present invention: The correlation between resource feature data and user learning interest feature data in a single screening of learning resource data is evaluated using text analysis and natural language processing techniques, including the following steps: The keyword information related to title, abstract, text content, and knowledge domain in the resource feature data of learning resource data is grouped, and the keyword information related to title, abstract, text content, and knowledge domain in the user's learning interest feature data is also grouped. The keywords related to title, abstract, text content, and knowledge domain in the resource feature data of learning resource data and the user's learning interest feature data are matched separately. Using the Word2Vec language model, keywords are mapped to a high-dimensional space to calculate the similarity of keywords in paired groups across titles, abstracts, text content, and knowledge domain data. Using the GloVe word vector model, we match the word groups in the paired groups of keywords by statistically analyzing the resource feature data of learning resource data and the user's learning interest feature data, and construct the corresponding word vectors based on the matched word groups. In evaluating the resource feature data of learning resource data and the learning interest feature data of users, the cosine similarity of the corresponding word vectors in the paired groups is calculated. The obtained cosine similarity is converted into a percentage, and the semantic matching degree of the keyword information in the paired groups is calculated. Based on the similarity between keywords in the paired groups and the degree of semantic matching of the corresponding data categories, the relevance evaluation value of the resource feature data of learning resource data and the learning interest feature data of users is calculated.
[0009] As a further aspect of the present invention: The cosine similarity of corresponding word vectors in paired groups is evaluated in the resource feature data of learning resource data and the learning interest feature data of users. The obtained cosine similarity is converted into a percentage, and the semantic matching degree of keyword information in paired groups is calculated, including the following steps: The cosine similarity of corresponding word vectors in paired groups within the resource feature data of learning resource data and the user's learning interest feature data is calculated using the following formula: Among them, COS(A1) i A2 i Let A1 be the cosine similarity of the j-th corresponding word vector. i and A2 i It is the vector dot product of the i-th corresponding word vector, |A1 i | and | A2 i | is the vector norm of the i-th corresponding word vector; The obtained cosine similarity is converted into a percentage, and the semantic matching degree of keyword information in the paired groups is calculated using the following formula: Where S represents the semantic matching degree of keyword information in the pairing group, i∈(1,2,…,n), and n is the total number of corresponding word vectors.
[0010] As a further aspect of the present invention: based on the similarity between keywords in the paired groups and the semantic matching degree of the corresponding data categories, the relevance evaluation value of the resource feature data of the learning resource data and the user's learning interest feature data is calculated, including the following steps: The TF-IDF algorithm was used to calculate the TF-IDF values of keywords in the original title, abstract, text content, and knowledge domain data text within the group. Based on the similarity of keywords in the paired groups and the TF-IDF values of the keywords, the overall keyword similarity score is calculated using the following formula: Where Z is the overall keyword similarity evaluation value in the paired groups, and T j Let X be the TF-IDF value of the j-th keyword. j Let be the similarity score of the j-th keyword, where j∈(1,2,…,m), and m is the total number of keywords in the group; Based on the overall similarity evaluation value of keywords in the paired groups and the semantic matching degree of keyword information, the relevance evaluation value of resource feature data of learning resource data and user learning interest feature data is calculated.
[0011] As a further aspect of the present invention: based on the overall similarity evaluation value of keywords in the paired groups and the semantic matching degree of keyword information, the relevance evaluation value of the resource feature data of the learning resource data and the user's learning interest feature data is calculated, including the following steps: Obtain the overall keyword similarity evaluation value and the semantic matching degree of keyword information in the calculated paired groups. Add the overall keyword similarity evaluation value and the semantic matching degree of keyword information in the paired groups to obtain the relevance evaluation value of the paired groups in the resource feature data of the learning resource data and the learning interest feature data of the user. Based on the relevance evaluation values of paired groups in the learning resource data and the user's learning interest data, different weights are assigned to the title, abstract, text content, and knowledge domain data groups, and then the values are summed to obtain the relevance evaluation values of the learning resource data and the user's learning interest data.
[0012] As a further aspect of the present invention: the correlation evaluation value between the resource feature data of learning resource data and the user's learning interest feature data is calculated using the following formula: Where E is the relevance evaluation value of resource feature data of learning resource data and user learning interest feature data, E1 is the relevance evaluation value of title keyword grouping, E2 is the relevance evaluation value of abstract keyword grouping, E3 is the relevance evaluation value of text content keyword grouping, and E4 is the relevance evaluation value of knowledge domain data keyword grouping.
[0013] As a further aspect of the present invention: based on the relevance evaluation results, the learning resource data selected in the first screening is further screened, and a recommended list of learning materials is constructed by selecting the learning resources, including the following steps: Based on the relevance evaluation results, the learning resource data selected in the first screening is further filtered to identify learning resources whose relevance evaluation value E between the resource feature data and the user's learning interest feature data is greater than or equal to 0.3 and less than or equal to 0.9. A learning material recommendation list is then constructed based on the selected learning resources.
[0014] As another aspect of the present invention: a learning resource recommendation system based on personalized recommendation, comprising: Behavioral data collection module: By analyzing the user's historical learning behavior, the module obtains the user's browsing history, collection history, and evaluation history. The browsing history includes: viewing history, click history, and viewing duration. User profile module: Extract related fields of user profile information by the fields of education and occupation information in the user profile information; extract the knowledge fields of learning materials corresponding to the user's browsing history, collection history and evaluation history as the browsing history related fields; Learning Feature Domain Construction Module: This module uses the user's professional domain, interest domain, user profile information association domain, and browsing history association domain as the user's learning feature domain; Learning resource acquisition module: Collects learning resource data, establishes a learning resource database, performs structured processing on the learning resource data, and extracts the characteristic information of the learning resource data, including: title, abstract, text content, and knowledge domain; Domain matching module: It filters the data in the learning resource database to select learning resources whose knowledge domain matches the user's learning characteristics domain; The relevance analysis module extracts resource feature data from learning resource data, extracts user learning interest feature data from user browsing history, and evaluates the relevance between resource feature data and user learning interest feature data in a single screening of learning resource data through text analysis and natural language processing techniques. Recommendation list building module: Based on the relevance evaluation results, the learning resource data selected in the first screening is further filtered, and a learning material recommendation list is built by selecting the learning resources.
[0015] Compared with the prior art, the beneficial effects of the present invention are: This invention constructs user profiles and analyzes user browsing history to facilitate personalized learning resource recommendations based on these profiles and browsing history. This ensures that the learning resources available to users better match their interests and learning needs, thereby improving learning effectiveness and satisfaction. By extracting features from user learning information, it facilitates a deeper understanding of users' learning habits, preferences, and behavioral patterns. Simultaneously, by analyzing the feature information of learning resources and extracting knowledge domain information, it facilitates faster location of the general domain of learning resources, saving search time and effort.
[0016] This invention filters learning resources from a database to identify those whose knowledge domains match the user's learning characteristics. It extracts resource feature data from the learning resources and user learning interest feature data from browsing history. Using text analysis and natural language processing (NLP) techniques, it evaluates the relevance between the resource feature data and the user's learning interest feature data from the initial screening. By ensuring the matching degree between the knowledge domains of the selected learning resources and the user's learning characteristics, it ensures that users access resources that better meet their learning needs and interests, improving learning efficiency and satisfaction. Combining the learning interest feature data extracted from user browsing history allows for a more accurate understanding of the user's learning preferences and interests. Furthermore, by utilizing text analysis and NLP techniques to evaluate the relevance between the resource features and learning interest features of the learning resources, it enables objective analysis and optimization of the learning resource recommendation mechanism. Attached Figure Description
[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art 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.
[0018] Figure 1 This is a schematic diagram of the method flow of the present invention; Figure 2 This is a schematic diagram of the method for extracting user learning interest feature data according to the present invention. Detailed Implementation
[0019] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0020] Please see Figures 1-2 The first aspect of this invention provides a method and system for recommending learning resources based on personalized recommendations, comprising the following steps: Collect users' learning information to construct user profiles, and extract learning feature domains based on user profiles and user browsing history; at the same time, extract feature information from learning resource data, and extract knowledge domain information from feature information; The data in the learning resource database is filtered to select learning resources whose knowledge domains match the user's learning characteristics. Extract resource feature data from learning resource data, extract user learning interest feature data from user browsing history, and evaluate the correlation between resource feature data and user learning interest feature data in a single screening of learning resource data through text analysis and natural language processing techniques. Based on the relevance evaluation results, the learning resource data selected in the first screening is further filtered, and a recommended list of learning materials is constructed by selecting the learning resources.
[0021] Furthermore, in this embodiment, user profile information is constructed by collecting user learning information, and learning feature domains are extracted based on user profile information and user browsing history. Simultaneously, feature information of learning resource data is extracted, along with knowledge domain information from the feature information. By constructing user profiles and analyzing user browsing history, personalized learning resource recommendations can be implemented based on these profiles and browsing history, ensuring that the learning resources obtained by users better match their interests and learning needs, thereby improving learning effectiveness and satisfaction. Extracting features from user learning information facilitates a deeper understanding of users' learning habits, preferences, and behavioral patterns. Furthermore, analyzing the feature information of learning resources and extracting knowledge domain information allows for faster location of the general domain of learning resources, saving search time and effort.
[0022] By filtering data in the learning resource database, learning resources whose knowledge domains match the user's learning characteristics are selected. Resource feature data is extracted from the learning resource data, and user learning interest feature data is extracted from user browsing history. Text analysis and natural language processing techniques are used to evaluate the relevance between the resource feature data and the user's learning interest feature data in the first round of filtering. By ensuring the matching degree between the knowledge domain of the selected learning resources and the user's learning characteristics, it is possible to ensure that users obtain resources that better meet their learning needs and interests, improving user learning efficiency and satisfaction. Combining the learning interest feature data extracted from user browsing history allows for a more accurate understanding of user learning preferences and interests. Using text analysis and natural language processing techniques to evaluate the relevance between resource features and learning interest features allows for objective analysis and optimization of the learning resource recommendation mechanism. By providing personalized learning resource recommendations and optimizing the learning experience, user satisfaction and experience quality can be significantly improved, not only increasing the platform's competitiveness but also helping to attract more users and improve user retention.
[0023] Based on the relevance evaluation results, the learning resource data selected in the first screening is further filtered, and a learning material recommendation list is constructed by selecting the learning resources. Using the relevance evaluation results for filtering helps ensure that the learning materials in the recommendation list have a high degree of matching with the user's interests, helping the user find the resources they really need more quickly. Filtering by relevance evaluation helps reduce the recommendation of irrelevant or low-relevance learning resources, thereby saving the user's time and energy, allowing them to focus more on learning itself rather than searching for materials.
[0024] In one embodiment of the present invention, user learning information is collected to construct user profile information, and learning feature domains are extracted based on user profile information and user browsing history; simultaneously, feature information of learning resource data is extracted, and knowledge domain information from the feature information is extracted, including the following steps: Collect learning information submitted by users, including: education level, occupation, professional field and interest field, to build user profile information; By analyzing users' historical learning behavior, we can obtain their browsing history, collection history, and evaluation history. The browsing history includes: viewing history, click history, and viewing duration. By analyzing the fields of education and occupation information in user profiles, we extract the fields associated with user profile information; we also extract the knowledge fields of learning materials corresponding to user browsing history, collection history, and evaluation history as the fields associated with browsing history. The user's professional field, interest field, user profile information association field, and browsing history association field are used as the user's learning feature fields; Collect learning resource data, establish a learning resource database, perform structured processing on the learning resource data, and extract the characteristic information of the learning resource data, including: title, abstract, text content, and knowledge domain.
[0025] In one embodiment of the present invention, the resource feature data of the learning resource data is extracted, and the user's learning interest feature data is extracted through the user's browsing history, including the following steps: Obtain feature information from learning resource data; Using the TextRank algorithm, we extract the title, summary, text content and keyword information in the knowledge domain from the feature information, and use the keyword information as the resource feature data of the learning resource data. We obtain feature information of corresponding learning materials from users' browsing history, collection history, and evaluation history. We then use the TextRank algorithm to extract keyword information of titles, abstracts, text content, and knowledge domains from the feature information of the corresponding learning materials, which serves as the user's learning interest feature data.
[0026] In one embodiment of the present invention, the correlation between resource feature data and user learning interest feature data in a first-selection learning resource data is evaluated using text analysis and natural language processing techniques, including the following steps: The keyword information related to title, abstract, text content, and knowledge domain in the resource feature data of learning resource data is grouped, and the keyword information related to title, abstract, text content, and knowledge domain in the user's learning interest feature data is also grouped. The keywords related to title, abstract, text content, and knowledge domain in the resource feature data of learning resource data and the user's learning interest feature data are matched separately. Using the Word2Vec language model, keywords are mapped to a high-dimensional space to calculate the similarity of keywords in paired groups across titles, abstracts, text content, and knowledge domain data. Using the GloVe word vector model, we match the word groups in the paired groups of keywords by statistically analyzing the resource feature data of learning resource data and the user's learning interest feature data, and construct the corresponding word vectors based on the matched word groups. In evaluating the resource feature data of learning resource data and the learning interest feature data of users, the cosine similarity of the corresponding word vectors in the paired groups is calculated. The obtained cosine similarity is converted into a percentage, and the semantic matching degree of the keyword information in the paired groups is calculated. Based on the similarity between keywords in the paired groups and the degree of semantic matching of the corresponding data categories, the relevance evaluation value of the resource feature data of learning resource data and the learning interest feature data of users is calculated.
[0027] Specifically, in this embodiment, using keyword similarity and semantic matching to evaluate resource feature data and user interest feature data can make the recommendation system more intelligent and accurate. The system can better understand users' true needs, thereby providing more targeted learning resource recommendations.
[0028] In one embodiment of the present invention, the cosine similarity of corresponding word vectors in paired groups is evaluated in the resource feature data of learning resource data and the learning interest feature data of users. The obtained cosine similarity is converted into a percentage, and the semantic matching degree of keyword information in paired groups is calculated, including the following steps: The cosine similarity of corresponding word vectors in paired groups within the resource feature data of learning resource data and the user's learning interest feature data is calculated using the following formula: Among them, COS(A1) i A2 i Let A1 be the cosine similarity of the j-th corresponding word vector. i and A2 i It is the vector dot product of the i-th corresponding word vector, |A1 i | and | A2 i | is the vector norm of the i-th corresponding word vector; The obtained cosine similarity is converted into a percentage, and the semantic matching degree of keyword information in the paired groups is calculated using the following formula: Where S represents the semantic matching degree of keyword information in the pairing group, i∈(1,2,…,n), and n is the total number of corresponding word vectors.
[0029] In one embodiment of the present invention, the relevance evaluation value of resource feature data of learning resource data and user learning interest feature data is calculated based on the similarity between keywords in the paired groups and the evaluated semantic matching degree of corresponding data categories, including the following steps: The TF-IDF algorithm was used to calculate the TF-IDF values of keywords in the original title, abstract, text content, and knowledge domain data text within the group. Based on the similarity of keywords in the paired groups and the TF-IDF values of the keywords, the overall keyword similarity score is calculated using the following formula: Where Z is the overall keyword similarity evaluation value in the paired groups, and T j Let X be the TF-IDF value of the j-th keyword. jLet be the similarity score of the j-th keyword, where j∈(1,2,…,m), and m is the total number of keywords in the group; Based on the overall similarity evaluation value of keywords in the paired groups and the semantic matching degree of keyword information, the relevance evaluation value of resource feature data of learning resource data and user learning interest feature data is calculated.
[0030] In one embodiment of the present invention, the relevance evaluation value of the resource feature data of the learning resource data and the learning interest feature data of the user is calculated based on the overall similarity evaluation value of keywords in the paired groups and the semantic matching degree of keyword information, including the following steps: Obtain the overall keyword similarity evaluation value and the semantic matching degree of keyword information in the calculated paired groups. Add the overall keyword similarity evaluation value and the semantic matching degree of keyword information in the paired groups to obtain the relevance evaluation value of the paired groups in the resource feature data of the learning resource data and the learning interest feature data of the user. Based on the relevance evaluation values of paired groups in the learning resource data and the user's learning interest data, different weights are assigned to the title, abstract, text content, and knowledge domain data groups, and then the values are summed to obtain the relevance evaluation values of the learning resource data and the user's learning interest data.
[0031] In one embodiment of the present invention, the correlation evaluation value between the resource feature data of learning resource data and the user's learning interest feature data is calculated using the following formula: Where E is the relevance evaluation value of resource feature data of learning resource data and user learning interest feature data, E1 is the relevance evaluation value of title keyword grouping, E2 is the relevance evaluation value of abstract keyword grouping, E3 is the relevance evaluation value of text content keyword grouping, and E4 is the relevance evaluation value of knowledge domain data keyword grouping.
[0032] In one embodiment of the present invention, the learning resource data selected in the first screening is further screened based on the relevance evaluation results, and a recommended list of learning materials is constructed by selecting the learning resources, including the following steps: Based on the relevance evaluation results, the learning resource data selected in the first screening is further filtered to identify learning resources whose relevance evaluation value E between the resource feature data and the user's learning interest feature data is greater than or equal to 0.3 and less than or equal to 0.9. A learning material recommendation list is then constructed based on the selected learning resources.
[0033] Specifically, in this embodiment, based on the relevance evaluation results—that is, the relevance evaluation value between the resource feature data of the learning resource data and the user's learning interest feature data—the learning resource data selected in the first screening is further screened. Through statistical analysis of a large amount of experimental data, it was found that when the relevance evaluation value E between the resource feature data of the learning resource data and the user's learning interest feature data is greater than or equal to 0.3, the learning resource has a certain degree of relevance to the user's browsing interests. However, when the relevance evaluation value E between the resource feature data of the learning resource data and the user's learning interest feature data is greater than 0.9, the content of the learning resource is too similar to the learning resources the user has browsed. To avoid duplication, when the relevance evaluation value E between the resource feature data of the learning resource data and the user's learning interest feature data is greater than 0.9, the corresponding learning resource is no longer recommended to the user.
[0034] Therefore, in this embodiment, the learning resource data selected in the first screening is further screened to select learning resources whose correlation evaluation value E between the resource feature data and the user's learning interest feature data is greater than or equal to 0.3 and less than or equal to 0.9. A learning material recommendation list is then constructed based on the selected learning resources.
[0035] As another embodiment of the present invention, a learning resource recommendation system based on personalized recommendation is provided, comprising: Behavioral data collection module: By analyzing the user's historical learning behavior, the module obtains the user's browsing history, collection history, and evaluation history. The browsing history includes: viewing history, click history, and viewing duration. User profile module: Extract related fields of user profile information by the fields of education and occupation information in the user profile information; extract the knowledge fields of learning materials corresponding to the user's browsing history, collection history and evaluation history as the browsing history related fields; Learning Feature Domain Construction Module: This module uses the user's professional domain, interest domain, user profile information association domain, and browsing history association domain as the user's learning feature domain; Learning resource acquisition module: Collects learning resource data, establishes a learning resource database, performs structured processing on the learning resource data, and extracts the characteristic information of the learning resource data, including: title, abstract, text content, and knowledge domain; Domain matching module: It filters the data in the learning resource database to select learning resources whose knowledge domain matches the user's learning characteristics domain; The relevance analysis module extracts resource feature data from learning resource data, extracts user learning interest feature data from user browsing history, and evaluates the relevance between resource feature data and user learning interest feature data in a single screening of learning resource data through text analysis and natural language processing techniques. Recommendation list building module: Based on the relevance evaluation results, the learning resource data selected in the first screening is further filtered, and a learning material recommendation list is built by selecting the learning resources.
[0036] The above embodiments are only used to illustrate the technical methods of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical methods of the present invention without departing from the spirit and scope of the technical methods of the present invention.
Claims
1. A learning resource recommendation method and system based on personalized recommendation, characterized in that, Includes the following steps: Collect users' learning information to build user profiles, and extract learning feature domains based on user profiles and user browsing history; Simultaneously, feature information is extracted from learning resource data, and knowledge domain information is extracted from the feature information; The data in the learning resource database is filtered to select learning resources whose knowledge domains match the user's learning characteristics. Extract resource feature data from learning resource data, extract user learning interest feature data from user browsing history, and evaluate the correlation between resource feature data and user learning interest feature data in a single screening of learning resource data through text analysis and natural language processing techniques. Based on the relevance evaluation results, the learning resource data selected in the first screening is further filtered, and a recommended list of learning materials is constructed by selecting the learning resources.
2. The learning resource recommendation method based on personalized recommendation according to claim 1, characterized in that, Collect users' learning information to build user profiles, and extract learning feature domains based on user profiles and user browsing history; Simultaneously, feature information is extracted from the learning resource data, and knowledge domain information is extracted from the feature information, including the following steps: Collect learning information submitted by users, including: education level, occupation, professional field and interest field, to build user profile information; By analyzing users' historical learning behavior, we can obtain their browsing history, collection history, and evaluation history. The browsing history includes: viewing history, click history, and viewing duration. By analyzing the fields of education and occupation information in user profiles, we extract the fields associated with user profile information; we also extract the knowledge fields of learning materials corresponding to user browsing history, collection history, and evaluation history as the fields associated with browsing history. The user's professional field, interest field, user profile information association field, and browsing history association field are used as the user's learning feature fields; Collect learning resource data, establish a learning resource database, perform structured processing on the learning resource data, and extract the characteristic information of the learning resource data, including: title, abstract, text content, and knowledge domain.
3. The learning resource recommendation method based on personalized recommendation according to claim 1, characterized in that, Extracting resource feature data from learning resource data and extracting user learning interest feature data from user browsing history includes the following steps: Obtain feature information from learning resource data; Using the TextRank algorithm, we extract the title, summary, text content and keyword information in the knowledge domain from the feature information, and use the keyword information as the resource feature data of the learning resource data. We obtain feature information of corresponding learning materials from users' browsing history, collection history, and evaluation history. We then use the TextRank algorithm to extract keyword information of titles, abstracts, text content, and knowledge domains from the feature information of the corresponding learning materials, which serves as the user's learning interest feature data.
4. The learning resource recommendation method based on personalized recommendation according to claim 3, characterized in that, Using text analysis and natural language processing techniques, the correlation between resource feature data and user learning interest feature data in a single screening of learning resource data is evaluated, including the following steps: The keyword information related to title, abstract, text content, and knowledge domain in the resource feature data of learning resource data is grouped, and the keyword information related to title, abstract, text content, and knowledge domain in the user's learning interest feature data is also grouped. The keywords related to title, abstract, text content, and knowledge domain in the resource feature data of learning resource data and the user's learning interest feature data are matched separately. Using the Word2Vec language model, keywords are mapped to a high-dimensional space to calculate the similarity of keywords in paired groups across titles, abstracts, text content, and knowledge domain data. Using the GloVe word vector model, we match the word groups in the paired groups of keywords by statistically analyzing the resource feature data of learning resource data and the user's learning interest feature data, and construct the corresponding word vectors based on the matched word groups. In evaluating the resource feature data of learning resource data and the learning interest feature data of users, the cosine similarity of the corresponding word vectors in the paired groups is calculated. The obtained cosine similarity is converted into a percentage, and the semantic matching degree of the keyword information in the paired groups is calculated. Based on the similarity between keywords in the paired groups and the degree of semantic matching of the corresponding data categories, the relevance evaluation value of the resource feature data of learning resource data and the learning interest feature data of users is calculated.
5. The learning resource recommendation method based on personalized recommendation according to claim 4, characterized in that, The evaluation process involves assessing the cosine similarity of corresponding word vectors in paired groups within resource feature data and user learning interest feature data. The cosine similarity is then converted to a percentage, and the semantic matching degree of keyword information within the paired groups is calculated. This includes the following steps: The cosine similarity of corresponding word vectors in paired groups within the resource feature data of learning resource data and the user's learning interest feature data is calculated using the following formula: Among them, COS(A1) i A2 i Let A1 be the cosine similarity of the j-th corresponding word vector. i and A2 i It is the vector dot product of the i-th corresponding word vector, |A1 i | and | A2 i | is the vector norm of the i-th corresponding word vector; The obtained cosine similarity is converted into a percentage, and the semantic matching degree of keyword information in the paired groups is calculated using the following formula: Where S represents the semantic matching degree of keyword information in the pairing group, i∈(1,2,…,n), and n is the total number of corresponding word vectors.
6. The learning resource recommendation method based on personalized recommendation according to claim 4, characterized in that, Based on the similarity between keywords in the paired groups and the semantic matching degree of the corresponding data categories, the relevance evaluation value of the resource feature data of the learning resource data and the user's learning interest feature data is calculated, including the following steps: The TF-IDF algorithm was used to calculate the TF-IDF values of keywords in the original title, abstract, text content, and knowledge domain data text within the group. Based on the similarity of keywords in the paired groups and the TF-IDF values of the keywords, the overall keyword similarity score is calculated using the following formula: Where Z is the overall keyword similarity evaluation value in the paired groups, and T j Let X be the TF-IDF value of the j-th keyword. j Let be the similarity score of the j-th keyword, where j∈(1,2,…,m), and m is the total number of keywords in the group; Based on the overall similarity evaluation value of keywords in the paired groups and the semantic matching degree of keyword information, the relevance evaluation value of resource feature data of learning resource data and user learning interest feature data is calculated.
7. The learning resource recommendation method based on personalized recommendation according to claim 4, characterized in that, Based on the overall keyword similarity evaluation value and the semantic matching degree of keyword information in the paired groups, the relevance evaluation value of the resource feature data of the learning resource data and the user's learning interest feature data is calculated, including the following steps: Obtain the overall keyword similarity evaluation value and the semantic matching degree of keyword information in the calculated paired groups. Add the overall keyword similarity evaluation value and the semantic matching degree of keyword information in the paired groups to obtain the relevance evaluation value of the paired groups in the resource feature data of the learning resource data and the learning interest feature data of the user. Based on the relevance evaluation values of paired groups in the learning resource data and the user's learning interest data, different weights are assigned to the title, abstract, text content, and knowledge domain data groups, and then the values are summed to obtain the relevance evaluation values of the learning resource data and the user's learning interest data.
8. The learning resource recommendation method based on personalized recommendation according to claim 7, characterized in that, The correlation score between resource feature data of learning resource data and user learning interest feature data is calculated using the following formula: Where E is the relevance evaluation value of resource feature data of learning resource data and user learning interest feature data, E1 is the relevance evaluation value of title keyword grouping, E2 is the relevance evaluation value of abstract keyword grouping, E3 is the relevance evaluation value of text content keyword grouping, and E4 is the relevance evaluation value of knowledge domain data keyword grouping.
9. The learning resource recommendation method based on personalized recommendation according to claim 8, characterized in that, Based on the relevance evaluation results, the learning resource data selected in the first screening is further filtered, and a recommended list of learning materials is constructed using the filtered learning resources. This includes the following steps: Based on the relevance evaluation results, the learning resource data selected in the first screening is further filtered to identify learning resources whose relevance evaluation value E between the resource feature data and the user's learning interest feature data is greater than or equal to 0.3 and less than or equal to 0.
9. A learning material recommendation list is then constructed based on the selected learning resources.
10. A learning resource recommendation system based on personalized recommendations, characterized in that, include: Behavioral data collection module: By analyzing the user's historical learning behavior, the module obtains the user's browsing history, collection history, and evaluation history. The browsing history includes: viewing history, click history, and viewing duration. User profile module: Extract related fields of user profile information by the fields of education and occupation information in the user profile information; extract the knowledge fields of learning materials corresponding to the user's browsing history, collection history and evaluation history as the browsing history related fields; Learning Feature Domain Construction Module: This module uses the user's professional domain, interest domain, user profile information association domain, and browsing history association domain as the user's learning feature domain; Learning resource acquisition module: Collects learning resource data, establishes a learning resource database, performs structured processing on the learning resource data, and extracts the characteristic information of the learning resource data, including: title, abstract, text content, and knowledge domain; Domain matching module: It filters the data in the learning resource database to select learning resources whose knowledge domain matches the user's learning characteristics domain; The relevance analysis module extracts resource feature data from learning resource data, extracts user learning interest feature data from user browsing history, and evaluates the relevance between resource feature data and user learning interest feature data in a single screening of learning resource data through text analysis and natural language processing techniques. Recommendation list building module: Based on the relevance evaluation results, the learning resource data selected in the first screening is further filtered, and a learning material recommendation list is built by selecting the learning resources.