Automatic training resource recommendation method and system

Abstract

The invention relates to the technical field of automatic training resource recommendation, in particular to an automatic training resource recommendation method and system, and the method comprises the steps: constructing a knowledge graph containing user and courseware information through a large language model (LLM); sorting the retrieval results by calculating the topology potential and the information entropy of the nodes in the knowledge graph; and determining recommended contents according to the sorting result. The method has the advantages that by integrating the multi-modal data processing capacity, comprehensive analysis and knowledge graph construction of various courseware contents such as texts, tables and videos and personnel information are achieved, and comprehensiveness and accuracy of information integration are improved; a dynamic association and real-time adaptation mechanism can adjust a graph structure according to user behaviors and historical data, so that the timeliness and correlation of recommended contents are ensured; the accurate recommendation method combining the topology potential and the entropy value improves the accuracy of the retrieval result, and better adapts to the complex scene requirements.

Description

Technical Field

[0001] The present invention relates to the technical field of automated training resource recommendation, and particularly to an automated training resource recommendation method and system. Background Art

[0002] Existing automated training resource recommendation systems mostly rely on single-modal data processing technologies, mainly performing resource matching and recommendation through independent feature extraction. These systems lack an effective integration mechanism when dealing with multi-modal data, resulting in recommendation results being limited to a single dimension and making it difficult to provide an all-round resource matching solution. At the same time, the semantic analysis of these systems for text mostly stays at keyword or shallow matching, making it difficult to understand deep semantic relationships and unable to accurately grasp the associations between complex contents, resulting in the lack of accuracy of the recommendation results. In addition, existing recommendation algorithms often compromise on accuracy when pursuing high computational performance, and vice versa, limiting the applicability of the system. Existing recommendation systems respond slowly to the dynamic changes of user interests, and the recommendation rules are relatively fixed, making it difficult to adjust the recommendation results according to the user's real-time behavior or environmental changes, resulting in a limited scope of application of the system. Summary of the Invention

[0003] The purpose of this part is to outline some aspects of the embodiments of the present invention and briefly introduce some preferred embodiments. Some simplifications or omissions may be made in this part, as well as in the abstract and title of the present application, to avoid obscuring the purpose of this part, the abstract, and the title. However, such simplifications or omissions shall not be used to limit the scope of the present invention.

[0004] In view of the above existing problems, the present invention is proposed.

[0005] Therefore, the present invention provides an automated training resource recommendation method and system, which can solve the problems mentioned in the background art.

[0006] To solve the above technical problems, the present invention provides the following technical solutions:

[0007] In a first aspect, the present invention provides an automated training resource recommendation method, including constructing a knowledge graph containing user and courseware information by using a large language model (LLM);

[0008] Sorting the retrieval results by calculating the topological potential and information entropy of the nodes in the knowledge graph;

[0009] Determining the recommended content according to the sorting results.

[0010] As a preferred solution of the automated training resource recommendation method of the present invention, wherein: constructing a knowledge graph containing user and courseware information by using a large language model (LLM) includes

[0011] Convert the user input into an optimized prompt;

[0012] Use a large language model (LLM) to parse the optimized prompt and generate a set of knowledge points;

[0013] Construct an information graph through the set of knowledge points, where the graph includes user nodes and courseware nodes, as well as edges representing the relationship between the user and the courseware.

[0014] As a preferred solution of the automated training resource recommendation method of the present invention, wherein: the retrieval results are sorted by calculating the topological potential and information entropy of the nodes in the knowledge graph, including

[0015] Evaluate the relevance by calculating the topological potential between the user and courseware nodes;

[0016] Evaluate the certainty by calculating the information entropy.

[0017] As a preferred solution of the automated training resource recommendation method of the present invention, wherein: the recommended content is determined according to the sorting result, including

[0018] Select the top k nodes with the highest scores in the sorting result as the recommended content;

[0019] Present the recommended content to the user in a visual or text form.

[0020] As a preferred solution of the automated training resource recommendation method of the present invention, wherein: the method includes dynamically adjusting the relevance threshold according to the user behavior and interaction data.

[0021] As a preferred solution of the automated training resource recommendation method of the present invention, wherein: the method further includes dynamically optimizing the graph structure and sorting algorithm according to the user's feedback on the recommended content.

[0022] As a preferred solution of the automated training resource recommendation method of the present invention, wherein: converting the user input into an optimized prompt includes

[0023] Convert the user input U = {u1, u2,..., un} into an optimized prompt P = f(U) through the prompt optimization function f.

[0024] In a second aspect, the present invention provides an automated training resource recommendation system, including: a construction module for constructing a knowledge graph containing user and courseware information by using a large language model (LLM);

[0025] A calculation and sorting module for sorting the retrieval results by calculating the topological potential and information entropy of the nodes in the knowledge graph;

[0026] A determination module for determining the recommended content according to the sorting result.

[0027] In a third aspect, the present invention provides a computer device, including a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, the steps of the above method are implemented.

[0028] In a fourth aspect, the present invention provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the above method are implemented.

[0029] Compared with the prior art, the beneficial effects of the present invention are as follows: By integrating multi-modal data processing capabilities, comprehensive analysis of various courseware contents such as text, tables, videos, etc., as well as personnel information, and the construction of a knowledge graph are realized, improving the comprehensiveness and accuracy of information integration; Its dynamic association and real-time adaptation mechanism can adjust the graph structure according to user behavior and historical data to ensure the timeliness and relevance of recommended content; The accurate recommendation method combining topological potential and entropy value improves the accuracy of retrieval results and better adapts to the requirements of complex scenarios; The feedback-driven self-optimization mechanism continuously optimizes the recommendation strategy through user feedback, continuously improving the recommendation quality and enhancing user satisfaction; At the same time, the knowledge graph construction framework of this solution has high flexibility and scalability, making it applicable not only to the field of education and training, but also to multiple scenarios such as medical care and enterprise management across different fields, showing broader applicability and innovation than the prior art. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for description in the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts. Among them:

[0031] Figure 1 It is a flowchart of an automated training resource recommendation method.

[0032] Figure 2 It is a schematic internal structure diagram of a computer device. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0033] To make the above objects, features, and advantages of the present invention more obvious and understandable, the specific embodiments of the present invention will be described in detail below with reference to the drawings in the specification.

[0034] Many specific details are set forth in the following description to provide a thorough understanding of the present invention. However, the present invention can also be implemented in other ways different from those described herein. Those skilled in the art can make similar extensions without departing from the connotation of the present invention. Therefore, the present invention is not limited by the specific embodiments disclosed below.

[0035] Secondly, the "one embodiment" or "embodiment" mentioned herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation manner of the present invention. The "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it an individual or alternative embodiment that is mutually exclusive with other embodiments.

[0036] Embodiment 1

[0037] Referring to Figure 1 , the first embodiment of the present invention provides an automated training resource recommendation method, which includes:

[0038] S1. Construct a knowledge graph containing user and courseware information using a large language model (LLM).

[0039] Furthermore, constructing a knowledge graph containing user and courseware information using a large language model (LLM) includes

[0040] Converting the user input into an optimized prompt.

[0041] Furthermore, converting the user input into an optimized prompt includes

[0042] Converting the user input U = {u1, u2,..., un} into an optimized prompt P = f(U) through the prompt optimization function f.

[0043] Parsing the optimized prompt using a large language model (LLM) to generate a set of knowledge points;

[0044] It should be noted that for text content extraction (for DOC and XLSX files, mainly processing text and table data): Assume that the file F contains N paragraphs of text or table cells:

[0045] F = {t1, t2,..., t N}.

[0046] Use the text parsing function to extract the text content:

[0047] C text = P doc (F),

[0048] where C text represents the extracted courseware content.

[0049] For table data, further analyze the cell relationships:

[0050] T tabular = {(r, c, v)|v is the value in the r-th row and c-th column of the table}.

[0051] Further, it should be noted that for video (MP4) data processing: Video V mp4 is decomposed into audio A and video frames F:

[0052] V mp4 → {A, F}.

[0053] The speech recognition model P is used for the audio part A andio to convert it into text:

[0054] C audio = P audio (A).

[0055] Image recognition is performed on the video frames F to extract scene and text information (OCR):

[0056] C frame = OCR(F) ∪ SceneDetection(F).

[0057] Integrate the audio and video frame information to form a complete representation of the courseware content:

[0058] C mp4 = C audio ∪ C frame .

[0059] Furthermore, it should be noted that for personnel information extraction in the management system: Personnel information is obtained from the management system and recorded in tabular form:

[0060] P = {(p1, a1), (p2, a2),..., (p m , a m )} where p i is the identifier of personnel a i and is its corresponding set of attributes (such as department, role, behavior history, etc.). Map the personnel information into vector form:

[0061] V(P) = {v1, v2,..., v m}

[0062] v i = Vectorize(a i ).

[0063] Construct an information graph through the knowledge point set, where the graph includes user nodes and courseware nodes, as well as edges representing the relationship between users and courseware.

[0064] It should be noted that the construction of the knowledge graph aims to present the courseware content, personnel information, and the relationship between the two in a structured form. The knowledge graph is represented in the form of nodes (entities) and edges (relationships), with weights assigned to the edges to reflect the degree of association between entities, and supports dynamic updates and queries.

[0065] Specifically, for node construction: Courseware nodes: These are the courseware contents extracted from multi-module data (text, table, and video data). Each node represents a courseware unit, such as a document, table, audio / video clip, etc.; User nodes: These are the personnel information extracted from the management system. Each node represents a user, with attached attributes such as role, behavior records, etc. Formula representation:

[0066] N = N C ∪N P ,

[0067] wherein, N c is the set of courseware nodes, and N p is the set of personnel nodes;

[0068] For edge construction: Edges represent the interactions (such as access, learning) between users and courseware or the associations (such as similarity) between courseware. Weight calculation:

[0069] w ij = α·sim(v i , v j ) + β·h ij ,

[0070]

[0071] wherein, sim(v i , v j ): The similarity between node feature vectors;

[0072] h ij : The weight influence of interaction frequency or historical behavior data;

[0073] α, β: Adjustment parameters that control the weights of similarity and behavior data.

[0074] Knowledge graph structure

[0075] Construct the overall structure G of the knowledge graph, which consists of a set of nodes N and a set of edges E, forming the knowledge graph G = (N, E). The dynamic weights of the nodes are adjusted through user behaviors and historical data, reflecting real-time and personalization:

[0076]

[0077] S2. Sort the retrieval results by calculating the topological potential and information entropy of the nodes in the knowledge graph.

[0078] Further, the retrieval results are sorted by calculating the topological potential and information entropy of the nodes in the knowledge graph, including

[0079] Evaluating the relevance by calculating the topological potential between the user and the courseware nodes;

[0080] Evaluating the certainty by calculating the information entropy.

[0081] It should be noted that the topological potential is used to evaluate the influence of a node in the knowledge graph, especially the relevance between the user and the courseware information. It reflects the connection strength between the node and its neighbor nodes and the overall position of its network structure. The topological potential of each node is equal to the sum of the connection weights between the node and all its neighbor nodes. The weight represents the similarity or relevance between nodes.

[0082]

[0083] According to the user behavior and historical data, the edge weights will be dynamically adjusted to make the graph reflect the real-time changes in user needs:

[0084]

[0085] The calculation results of the topological potential are used to sort and filter nodes to identify the nodes that contribute most to the user's needs.

[0086] Furthermore, it should be noted that the entropy value is used to evaluate the uncertainty of node information. A lower entropy value indicates that the information is more certain and relevant, which is suitable for recommendation or retrieval. The entropy value of node information is calculated based on the information probability distribution p i Calculated, the entropy value formula is as follows:

[0087]

[0088] The probability distribution p i Can be obtained by normalizing the weights of the nodes:

[0089]

[0090] Through the evaluation of the entropy value, nodes with high information certainty can be identified, and nodes with low relevance or fuzzy information to the user's needs can be filtered out.

[0091] FilteredNodes = {i|H(i) < Threshold} S3. Determine the recommended content according to the sorting results.

[0092] Further, according to the sorting results, determine the recommended content, including

[0093] Select the k nodes with the highest scores in the sorting results as the recommended content;

[0094] Present the recommended content to the user in a visual or text form.

[0095] It should be noted that based on the weight evaluation of topological potential and entropy value, the nodes are sorted by relevance. The objective function comprehensively considers the topological potential and entropy value of the nodes and is defined as follows:

[0096] S = λ1·φ - λ2·H

[0097] Sorting rule: The sorting result SSS of the nodes is jointly determined by the contribution of the topological potential φ and the suppression of the entropy value H. The weight parameters λ1 and λ2 are used to adjust the influence ratio of the two.

[0098] Obtain the recommended result: According to the sorted score S, select the top k nodes with the highest scores as the recommended content:

[0099] C rec = {c1, c2,..., c k}, k = argmax k S(c k )

[0100] Recommendation output: Present the filtered recommended content to the user in a visual or text form, ensuring the relevance and intuitiveness of the result, and at the same time dynamically update the knowledge graph to adapt to the user's continuous interaction.

[0101] Furthermore, the method includes dynamically adjusting the relevance threshold according to the user's behavior and interaction data.

[0102] It should be noted that to improve the accuracy of the recommended result, a dynamic threshold adjustment strategy is adopted, that is, according to the user's real-time behavior and interaction data, the relevance threshold for filtering nodes is updated. Specifically, let the initial threshold be T0, and the adjustment formula is as follows:

[0103] T new = T0 + η·(FeedbackScore - ExpectedScore)

[0104] η is the learning rate, which controls the amplitude of the threshold adjustment.

[0105] FeedbackScore is the user's score or click-through rate for the recommended result.

[0106] ExpectedScore is the expected score of the system for the result

[0107] Furthermore, the method also includes dynamically optimizing the graph structure and sorting algorithm according to the user's feedback on the recommended content.

[0108] It should be noted that dynamically optimizing the graph structure and sorting algorithm specifically includes:

[0109]

[0110] FeedbackImpact represents the impact weight of user behaviors such as clicks and dwell time.

[0111] The parameters α and β control the balance between the original weight and the feedback correction.

[0112] Continuous iteration: Through multiple rounds of interaction and feedback optimization, gradually improve the recommendation system's understanding of user needs, ensure that the system can dynamically adapt to user changes, and achieve continuous improvement of recommendation results.

[0113] In summary, the beneficial effects of the automated training resource recommendation method of the present invention are as follows: By integrating multi-modal data processing capabilities, it realizes the comprehensive analysis of various courseware contents such as text, tables, and videos, as well as personnel information, and constructs a knowledge graph, improving the comprehensiveness and accuracy of information integration; Its dynamic association and real-time adaptation mechanism can adjust the graph structure according to user behaviors and historical data, ensuring the timeliness and relevance of recommended content; The precise recommendation method combining topological potential and entropy value improves the accuracy of retrieval results and better adapts to the needs of complex scenarios; The feedback-driven self-optimization mechanism continuously optimizes the recommendation strategy through user feedback, continuously improves the recommendation quality, and enhances user satisfaction; At the same time, the knowledge graph construction framework of this solution has high flexibility and scalability, making it not only applicable to the field of education and training, but also applicable to multiple scenarios such as medical care and enterprise management across domains, demonstrating broader applicability and innovation than existing technologies.

[0114] Embodiment 2

[0115] This embodiment provides an automated training resource recommendation system, which includes a construction module for constructing a knowledge graph containing user and courseware information using a large language model (LLM);

[0116] A calculation and sorting module for sorting retrieval results by calculating the topological potential and information entropy of nodes in the knowledge graph;

[0117] A determination module for determining recommended content according to the sorting results.

[0118] The above-mentioned unit modules can be embedded in or independent of the processor in the computer device in hardware form, or stored in the memory of the computer device in software form, so that the processor can call and execute the operations corresponding to the above modules.

[0119] Embodiment 3

[0120] This embodiment provides a computer device, which can be a terminal, and its internal structure diagram can be as Figure 2As shown. The computer device includes a processor, a memory, a communication interface, a display screen, and an input device connected via a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The communication interface of the computer device is used to communicate with external terminals in a wired or wireless manner, and the wireless manner can be achieved through WIFI, carrier network, NFC (Near Field Communication), or other technologies. When the computer program is executed by the processor, it implements an automated training resource recommendation method. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covering the display screen, or buttons, trackballs, or touchpads provided on the computer device housing, or an external keyboard, touchpad, or mouse, etc.

[0121] This embodiment also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it realizes: constructing a knowledge graph containing user and courseware information by using a large language model (LLM); sorting the retrieval results by calculating the topological potential and information entropy of nodes in the knowledge graph; determining the recommended content according to the sorting results.

[0122] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention, and they should all be covered within the scope of the claims of the present invention.

Claims

1. An automated training resource recommendation method, characterized in that: including constructing a knowledge graph containing user and courseware information by using a large language model (LLM); sorting the retrieval results by calculating the topological potential and information entropy of nodes in the knowledge graph; determining the recommended content according to the sorting results.

2. The automated training resource recommendation method according to claim 1, wherein: The constructing a knowledge graph containing user and courseware information by using a large language model (LLM) includes converting the user input into an optimized prompt; using the large language model (LLM) to parse the optimized prompt to generate a set of knowledge points; constructing an information graph through the set of knowledge points, where the graph includes user nodes and courseware nodes, and edges representing the relationship between users and courseware.

3. The automated training resource recommendation method according to claim 2, wherein: The sorting the retrieval results by calculating the topological potential and information entropy of nodes in the knowledge graph includes evaluating the relevance by calculating the topological potential between user and courseware nodes; evaluating the certainty by calculating the information entropy.

4. The automated training resource recommendation method according to claim 3, characterized in that: The determining the recommended content according to the sorting results includes selecting the top k nodes with the highest scores in the sorting results as the recommended content; presenting the recommended content to the user in a visual or text form.

5. The automated training resource recommendation method according to claim 4, wherein: The method includes dynamically adjusting the relevance threshold according to user behavior and interaction data.

6. The automated training resource recommendation method according to claim 5, wherein: The method further includes dynamically optimizing the graph structure and sorting algorithm according to the user's feedback on the recommended content.

7. The automated training resource recommendation method according to any one of claims 2 to 6, characterized in that: The converting the user input into an optimized prompt includes converting the user input U = {u1, u2,..., un} into an optimized prompt P = f(U) through the prompt optimization function f.

8. An automated training resource recommendation system, characterized in that, including: a construction module for constructing a knowledge graph containing user and courseware information by using a large language model (LLM); a calculation and sorting module for sorting the retrieval results by calculating the topological potential and information entropy of nodes in the knowledge graph; a determination module for determining the recommended content according to the sorting results.

9. A computer device, comprising a memory and a processor, the memory storing a computer program, characterized in that, When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images