Children English content personalized recommendation method and system based on multi-feature fusion and hierarchical recall strategy

Abstract

The invention relates to the technical field of child English learning artificial intelligence recommendation systems, and discloses a child English content personalized recommendation method and system based on multi-feature fusion and a hierarchical recall strategy, and the method comprises the following steps: S1, constructing a multi-dimensional feature model which comprises the natural attributes, cognitive levels and behavior preferences of a user side, vocabulary difficulty grading, theme classification and freshness of the content side are obtained; and S2, implementing a hierarchical recall strategy, and generating a candidate content pool through a high-heat model, ICF recall based on content attributes and UCF recall based on similar user group behaviors in sequence. Through a multi-feature fusion technology, multi-dimensional features such as the age, the English level, the learning interest and the historical behavior of the child user are comprehensively considered, more accurate personalized recommendation is achieved, compared with a traditional recommendation system, user requirements can be more comprehensively understood, the English content better conforming to the individual difference of children is provided, and the user experience is improved. Therefore, the learning interest and efficiency of children are obviously improved.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence recommendation systems for children's English learning, specifically to a personalized recommendation method and system for children's English content based on multi-feature fusion and hierarchical recall strategies. Background Technology

[0002] With the rapid development of artificial intelligence technology, recommender systems have been widely used in various fields. However, in the field of AI recommender systems for children's English learning, existing systems still have many shortcomings.

[0003] Traditional collaborative filtering recommendation methods, such as user collaborative filtering (UCF) and item collaborative filtering (ICF), primarily rely on the similarity of individual users or content for recommendations. This approach fails to adequately consider the differences in cognitive abilities resulting from children's natural attributes (such as age and gender), leading to recommended content that may be beyond or below their cognitive level. Furthermore, new users or content are difficult to match accurately due to data sparsity, resulting in a significant cold start problem. In addition, the recommendation results rely excessively on users' historical behavior, leading to insufficient content diversity.

[0004] On the other hand, the fixed content pool recommendation model dominated by human editors also has obvious drawbacks. Content updates are lagging behind, failing to dynamically adapt to changes in children's user preferences. Furthermore, because content attributes (such as vocabulary difficulty and topic freshness) are not quantified, the recommendation results lack scientific basis and are difficult to meet the personalized needs of AI recommendation systems for children's English learning.

[0005] Therefore, the present invention aims to solve the above-mentioned problems in the prior art and provide a personalized recommendation method and system for children's English content based on multi-feature fusion and hierarchical recall strategy, so as to achieve more accurate, diversified and personalized recommendations. Summary of the Invention

[0006] The purpose of this invention is to provide a personalized recommendation method and system for children's English content based on multi-feature fusion and hierarchical recall strategy, so as to solve the problems mentioned in the background art.

[0007] To achieve the above objectives, the present invention provides the following technical solution: a personalized recommendation method for children's English content based on multi-feature fusion and hierarchical recall strategy, comprising the following steps: Step S1: Construct a multi-dimensional feature model, including natural attributes, cognitive level and behavioral preferences on the user side, and vocabulary difficulty level, topic classification and freshness on the content side; Step S2: Implement a hierarchical recall strategy, generating a candidate content pool by sequentially using a high-heat model, ICF recall based on content attributes, and UCF recall based on similar user group behavior. Step S3: Employ a dynamic weight adjustment mechanism to assign different weights to content freshness, theme similarity, user attribute suitability, and cognitive matching based on user age and cognitive level, and calculate the final score. Step S4: Sort the candidate content according to the final score and generate a personalized recommendation list.

[0008] Preferably, the user-side natural attributes include age and gender, cognitive level is assessed through test scores and learning progress, and behavioral preferences are recorded through user clicks, skips, and repetitions of content.

[0009] Preferably, the vocabulary difficulty level of the content side adopts the CEFR standard, and the topic classification includes, but is not limited to, animals, family, numbers, etc., and the freshness is calculated by the publication time decay factor.

[0010] Preferably, in the hierarchical recall strategy, the high-popularity model selects high-popularity content on the current platform as the basic candidate pool, the ICF recall expands similar content based on content attribute similarity, and the UCF recall supplements long-tail content based on similar user group behavior data.

[0011] Preferably, in the dynamic weight adjustment mechanism, a higher freshness weight is assigned to younger users (e.g., 0-3 years old), and a higher cognitive matching weight is assigned to older users (e.g., 8-12 years old).

[0012] Preferably, the final score calculation formula is: Final Score = Popularity Decay Factor × Content Similarity + User Attribute Adaptability × Cognitive Matching.

[0013] Another technical problem to be solved by this invention is to provide a system for a personalized recommendation method of children's English content based on the multi-feature fusion and hierarchical recall strategy described above, comprising: The feature engineering module is used to clean user attributes, behavior logs, and content tags in real time. The recall engine is used to execute the three-layer recall of high-hot model, ICF and UCF in parallel to generate a candidate content pool; A ranking model used to reorder candidate content based on a dynamic weight formula; A feedback mechanism is used to update feature weights based on real-time user interaction.

[0014] Preferably, the feature engineering module further includes a user feature extraction submodule and a content feature extraction submodule, which are used to extract multi-dimensional features from the user side and the content side, respectively.

[0015] Preferably, the recall engine uses a parallel processing mechanism to efficiently execute a three-layer recall strategy, ensuring the diversity and personalization of recommended content.

[0016] Preferably, the ranking model generates the final recommendation list by comprehensively considering content popularity, similarity, user attribute suitability, and cognitive matching degree according to the dynamic weight formula.

[0017] This invention provides a method and system for personalized recommendation of English content for children based on multi-feature fusion and hierarchical recall strategies. It has the following beneficial effects: (1) This invention uses multi-feature fusion technology to comprehensively consider the multi-dimensional characteristics of children's users, such as age, English proficiency, learning interest, and historical behavior, to achieve more accurate personalized recommendations. Compared with traditional recommendation systems, this invention can more comprehensively understand user needs and provide English content that is more in line with the individual differences of children, thereby significantly improving children's learning interest and efficiency.

[0018] (2) This invention targets new users or users who lack historical behavioral data (i.e., the cold start problem). This invention uses a hierarchical recall strategy, which combines the basic attributes of users (such as age and English proficiency) and the basic attributes of content (such as difficulty and topic), to perform preliminary content screening and recommendation. This method can still provide relatively reasonable recommendation results even when user behavioral data is insufficient, effectively alleviate the cold start problem, and ensure smooth and efficient recommendation.

[0019] (3) In the recommendation process, this invention not only considers personalized needs but also emphasizes the diversity of content. Through a hierarchical recall strategy, it first ensures broad coverage of recommended content, and then conducts refined screening through multi-feature fusion. Thus, while meeting personalized needs, it also ensures that child users can access diverse English content. This balance helps promote children's all-round development and improve their overall English learning effect. Attached Figure Description

[0020] Figure 1 This is a schematic diagram of the method steps of the present invention; Figure 2 This is a system architecture view of the present invention. Detailed Implementation

[0021] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. 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.

[0022] Examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain the invention, and should not be construed as limiting the invention.

[0023] A preferred embodiment of the personalized recommendation method and system for children's English content based on multi-feature fusion and hierarchical recall strategy provided by this invention is as follows: Figure 1-2 As shown: A personalized recommendation method for children's English content based on multi-feature fusion and hierarchical recall strategy includes the following steps: Step S1: Construct a multi-dimensional feature model, including natural attributes, cognitive level and behavioral preferences on the user side, and vocabulary difficulty level, topic classification and freshness on the content side; User-side natural attributes include age and gender, cognitive level is assessed through test scores and learning progress, and behavioral preferences are recorded through user clicks, skips, and repetitive actions on content.

[0024] The vocabulary difficulty level of the content is based on the CEFR standard, and the subject categories include, but are not limited to, animals, family, numbers, etc. Freshness is calculated using the publication time decay factor.

[0025] Step S2: Implement a hierarchical recall strategy, generating a candidate content pool by sequentially using a high-heat model, ICF recall based on content attributes, and UCF recall based on similar user group behavior. In the hierarchical recall strategy, the high-popularity model selects the most popular content on the current platform as the basic candidate pool, the ICF recall expands similar content based on content attribute similarity, and the UCF recall supplements long-tail content based on similar user group behavior data.

[0026] Step S3: Employ a dynamic weight adjustment mechanism to assign different weights to content freshness, theme similarity, user attribute suitability, and cognitive matching based on user age and cognitive level, and calculate the final score. In the dynamic weight adjustment mechanism, a higher freshness weight is assigned to younger users (e.g., 0-3 years old), and a higher cognitive matching weight is assigned to older users (e.g., 8-12 years old).

[0027] The final score is calculated as follows: Final Score = Popularity Decay Factor × Content Similarity + User Attribute Adaptability × Cognitive Match.

[0028] Step S4: Sort the candidate content according to the final score and generate a personalized recommendation list.

[0029] Another technical problem this invention aims to solve is to provide a personalized recommendation system for children's English content based on multi-feature fusion and hierarchical recall strategies, including: The feature engineering module is used to clean user attributes, behavior logs, and content tags in real time. The recall engine is used to execute the three-layer recall of high-hot model, ICF and UCF in parallel to generate a candidate content pool; A ranking model used to reorder candidate content based on a dynamic weight formula; A feedback mechanism is used to update feature weights based on real-time user interaction.

[0030] The feature engineering module further includes a user feature extraction submodule and a content feature extraction submodule, which are used to extract multi-dimensional features from the user side and the content side, respectively.

[0031] The recall engine uses a parallel processing mechanism to efficiently execute a three-layer recall strategy, ensuring the diversity and personalization of recommended content.

[0032] The ranking model generates the final recommendation list by comprehensively considering content popularity, similarity, user attribute suitability, and cognitive matching based on a dynamic weight formula.

[0033] Taking a 3-year-old girl as an example, the recommendation process of this invention will be explained in detail: Data input: Natural attributes: Age 3, Gender: Female; Behavioral data: Recently, the child frequently clicked on "animal-themed" nursery rhymes and skipped the "number cognition" animation; Cognitive level: The test showed that the mastery of "color vocabulary" reached 90%, and "shape vocabulary" reached 60%.

[0034] Hierarchical Recall and Ranking: First layer (high-popularity model): Select the top 10 "animal-themed" nursery rhymes from the high-popularity content pool; The second layer (ICF recall): expand similar content, such as the "zoo scene dialogue" animation (vocabulary difficulty Lv1, theme similarity 0.85). The third layer (UCF recall): Supplement with "family daily interaction" videos (long-tail content, user group click-through rate >70%) based on similar user group data.

[0035] Dynamic weight calculation: Since the user's age is ≤3 years old, the weighting is set as follows: freshness 60% and cognitive matching 40%. Calculations for the "Zoo Scene Dialogue" animation: Score = 0.8 (Popularity Decay) × 0.9 (Topic Similarity) + 0.6 (Age Match) × 0.7 (Vocabulary Fit) = 1.26 Generate a recommendation list: The final recommended list is generated by sorting by score, including the "Zoo Scene Dialogue" animation and other high-scoring content.

[0036] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0037] Finally, it should be noted that the above descriptions are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A personalized recommendation method for children's English content based on multi-feature fusion and hierarchical recall strategy, characterized in that, Includes the following steps: Step S1: Construct a multi-dimensional feature model, including natural attributes, cognitive level and behavioral preferences on the user side, and vocabulary difficulty level, topic classification and freshness on the content side; Step S2: Implement a hierarchical recall strategy, generating a candidate content pool by sequentially using a high-heat model, ICF recall based on content attributes, and UCF recall based on similar user group behavior. Step S3: Employ a dynamic weight adjustment mechanism to assign different weights to content freshness, theme similarity, user attribute suitability, and cognitive matching based on user age and cognitive level, and calculate the final score. Step S4: Sort the candidate content according to the final score and generate a personalized recommendation list.

2. The personalized recommendation method and system for children's English content based on multi-feature fusion and hierarchical recall strategy according to claim 1, characterized in that: The user's natural attributes include age and gender, cognitive level is assessed through test scores and learning progress, and behavioral preferences are recorded through user clicks, skips, and repetitions of content.

3. The personalized recommendation method for children's English content based on multi-feature fusion and hierarchical recall strategy according to claim 1, characterized in that: The vocabulary difficulty level of the content side adopts the CEFR standard, and the topic classification includes, but is not limited to, animals, family, numbers, etc. Freshness is calculated by the publication time decay factor.

4. The personalized recommendation method for children's English content based on multi-feature fusion and hierarchical recall strategy according to claim 1, characterized in that: In the hierarchical recall strategy, the high-popularity model selects high-popularity content on the current platform as the basic candidate pool, the ICF recall expands similar content based on content attribute similarity, and the UCF recall supplements long-tail content based on similar user group behavior data.

5. The personalized recommendation method for children's English content based on multi-feature fusion and hierarchical recall strategy according to claim 1, characterized in that: In the dynamic weight adjustment mechanism, a higher freshness weight is assigned to younger users (e.g., 0-3 years old), and a higher cognitive matching weight is assigned to older users (e.g., 8-12 years old).

6. The personalized recommendation method for children's English content based on multi-feature fusion and hierarchical recall strategy according to claim 1, characterized in that: The final score is calculated as follows: Final Score = Popularity Decay Factor × Content Similarity + User Attribute Adaptability × Cognitive Match.

7. A system for personalized recommendation of children's English content based on the multi-feature fusion and hierarchical recall strategy described in claims 1-6, characterized in that: include: The feature engineering module is used to clean user attributes, behavior logs, and content tags in real time. The recall engine is used to execute the three-layer recall of high-hot model, ICF and UCF in parallel to generate a candidate content pool; A ranking model used to reorder candidate content based on a dynamic weight formula; A feedback mechanism is used to update feature weights based on real-time user interaction.

8. The personalized recommendation system for children's English content based on multi-feature fusion and hierarchical recall strategy according to claim 7, characterized in that: The feature engineering module further includes a user feature extraction submodule and a content feature extraction submodule, which are used to extract multi-dimensional features from the user side and the content side, respectively.

9. The personalized recommendation system for children's English content based on multi-feature fusion and hierarchical recall strategy according to claim 7, characterized in that: The recall engine uses a parallel processing mechanism to efficiently execute a three-layer recall strategy, ensuring the diversity and personalization of recommended content.

10. The personalized recommendation system for children's English content based on multi-feature fusion and hierarchical recall strategy according to claim 7, characterized in that: The ranking model generates a final recommendation list by comprehensively considering content popularity, similarity, user attribute suitability, and cognitive matching based on a dynamic weight formula.

Images