Method, system and terminal for generating employment skill learning path based on machine question answering

Abstract

This invention belongs to the field of skills training technology, specifically relating to a method, system, and terminal for generating employment skills learning paths based on machine question answering. It includes generating an initial formal learning path; and dynamically adjusting the learning path in response to the detection that the user's current learning progress duration is less than the learning termination condition set for the formal learning path. The dynamic adjustment of the learning path includes determining a learning bias direction based on the user's answer records in the formal learning path; and selecting supplementary learning content from a preset set of learning content based on the learning bias direction to update the formal learning path. This invention enhances the personalization of the path by evaluating the user's historical answer process and learning status parameters, achieves iterative optimization of the path through closed-loop feedback of adjusting a second preset model based on recommendation accuracy scores, and ensures the integrity of learning through dynamic adjustment of the learning path, thereby improving the accuracy of learning content recommendations and the overall learning effect.

Description

Technical Field

[0001] This invention belongs to the field of employment skills training technology, specifically relating to a method, system, and terminal for generating employment skills learning paths based on machine question answering. Background Technology

[0002] With the continuous development of information technology, online education and vocational training systems have become a key component of personal skills enhancement and corporate talent development. In order to improve learning efficiency and meet the personalized needs of different users, it is particularly important to build an adaptive learning system that can intelligently plan learning content. Such systems aim to optimize the learning process and improve the depth and speed of knowledge acquisition by recommending customized learning paths to users. They have significant application value in today's rapidly iterating knowledge society.

[0003] However, existing personalized learning path generation technologies suffer from prominent static and unidirectional problems in path planning. Most systems generate a fixed learning path only before learning begins, based on pre-defined logical dependencies between knowledge points. They lack dynamic adjustment mechanisms throughout the learning process, failing to adapt to changes in users' cognitive levels and interests. Furthermore, user feedback mechanisms are lacking and limited in scope. Existing technologies rarely consider or effectively utilize user behavioral feedback during the learning process, particularly ignoring learning preferences revealed when users choose review materials. For example, the system cannot determine whether a user prefers theoretical review, case analysis, or practical exercises, thus failing to verify whether the recommended learning content and related review materials are truly suitable for the user. In addition, the path's effectiveness verification and iteration capabilities are severely inadequate. Without acquiring and analyzing user preferences for review materials, the rationality of the current learning path cannot be verified, and subsequent paths cannot be intelligently optimized and adjusted based on this implicit feedback. These problems often result in generated learning paths that are disconnected from users' actual needs and learning habits, easily leading to user resistance and ultimately interfering with learning efficiency and effectiveness.

[0004] To address the aforementioned issues, this invention proposes a method, system, and terminal for generating employment skills learning paths based on machine question answering. Summary of the Invention

[0005] The purpose of this invention is to provide a method, system, and terminal for generating employment skills learning paths based on machine question answering. This method can verify the review materials in the learning path, output the review direction of the review materials, thereby updating the learning path and recommending suitable learning materials to users.

[0006] The technical solution adopted in this invention is as follows: a method for generating employment skills learning paths based on machine question answering, comprising the following steps:

[0007] Based on the user's historical answer process and results, an initial formal learning path is generated;

[0008] And when the user's current learning progress time is detected to be less than the learning termination condition set for the formal learning path, dynamic adjustment of the learning path is executed, which includes:

[0009] Based on the user's answer records in the formal learning path, determine the learning bias direction;

[0010] Based on the learning preferences, supplementary learning content is selected from all the preset learning content to update the formal learning path;

[0011] Among these, determining the learning bias direction based on the user's answer records in the formal learning path includes:

[0012] After the user completes the learning and answers questions according to the formal learning path, the user's answer accuracy rate is obtained to generate a post-learning score;

[0013] The learning bias score is calculated based on the user's answer records for each knowledge node completed in the formal learning path.

[0014] A learning bias score is generated by combining the learning bias score with the score after learning.

[0015] The learning bias score is input into the second preset model to output recommended learning content;

[0016] Based on all the preset learning content, the distribution of recommended learning content within the preset learning content is analyzed to determine the direction of learning bias.

[0017] Preferably, the initial formal learning path is generated based on the user's historical answering process and results, including:

[0018] The historical response process is processed to generate response preference tags;

[0019] Based on historical responses, calculate the user's response score as a learning status parameter;

[0020] Based on answer preference tags and learning state parameters, an initial formal learning path is generated through a path generation function.

[0021] Preferably, the answer preference tags include answer time tags and answer accuracy tags;

[0022] The process of processing historical responses to generate answer preference tags includes:

[0023] Extract the answering time and answering result from the historical answering process;

[0024] Calculate the average answering time based on the answering time to generate answering time tags;

[0025] The accuracy rate label is generated by weighting the answers with the pre-set complexity of the corresponding test questions.

[0026] Preferably, the user's answer score includes a score to be verified and a reference score;

[0027] Also includes:

[0028] The user's overall learning score is calculated by combining the score to be verified with the reference score using a scoring function.

[0029] The determination of learning status parameters is further based on the user's overall learning score.

[0030] Preferably, the learning termination condition is set based on the estimated total duration of the formal learning path;

[0031] Based on learning preferences, supplementary learning content is selected from the pre-set learning materials to update the formal learning path, including:

[0032] Calculate the current learning progress duration and the learning termination conditions to determine the remaining learning time;

[0033] Supplementary learning content is added to the formal learning path, where the selection of supplementary learning content is constrained by the remaining learning time.

[0034] Preferably, the method further includes:

[0035] Obtain the actual review content selected by the user after receiving the recommended learning content, as the user's actual review path;

[0036] The recommended learning content is compared with the user's actual review path to calculate the overlap between the two contents, and a recommendation accuracy score is generated based on the overlap.

[0037] The second preset model is adjusted based on the recommendation accuracy score.

[0038] This invention also discloses a machine question-answering-based employment skills learning path generation system, comprising the following modules:

[0039] The user status acquisition module is used to acquire the user's historical answering process and historical answering results;

[0040] The learning path generation module is used to generate an initial formal learning path based on the historical answering process and historical answering results;

[0041] The learning effectiveness evaluation module is used to evaluate the learning effectiveness after the user learns along the formal learning path to generate a learning bias score and a recommendation accuracy score.

[0042] The path dynamic adjustment module, in response to the detection that the user's current learning progress time is less than the learning termination condition set for the formal learning path, is configured as follows:

[0043] Based on the learning bias score generated by the learning effectiveness evaluation module, the direction of learning bias is determined;

[0044] Supplementary learning content is selected based on learning preferences to update the formal learning path;

[0045] The system is also configured to adjust the second preset model based on the recommendation accuracy score generated by the learning effect evaluation module.

[0046] Preferably, the initial formal learning path is generated based on the historical answering process and historical answering results, including:

[0047] Process historical responses to generate answer preference tags;

[0048] Based on historical responses, calculate the user's response score as a learning status parameter;

[0049] An initial formal learning path is generated based on answer preference tags and learning status parameters.

[0050] Preferably, evaluating learning effectiveness to generate learning bias scores and recommendation accuracy scores includes:

[0051] Obtain the accuracy rate of user responses to generate a post-learning score;

[0052] Based on the user's answer records for each knowledge node completed in the formal learning path, a learning bias score is calculated, and then the post-learning score is combined with the learning bias score to generate a learning bias rating.

[0053] Obtain the user's actual review path and compare it with the recommended learning content to generate a recommendation accuracy score.

[0054] The present invention also discloses a terminal for generating employment skills learning paths based on machine question answering, including a processor and a memory connected in communication with the processor. The memory stores a computer program, and when the computer program is executed by the processor, it implements a method for generating employment skills learning paths based on machine question answering.

[0055] Beneficial effects

[0056] This invention obtains a user's historical answering process and results, generates answer preference tags representing the user's answering habits and learning state parameters representing the level of knowledge mastery, and inputs both into a path generation function to output an initial formal learning path. This allows for a comprehensive evaluation based on both the user's answering behavior and knowledge mastery level. Compared to existing technologies that rely solely on knowledge scores, the generated initial formal learning path is more closely aligned with the user's individual learning style and cognitive characteristics, improving the personalization and initial matching degree of the learning path, thereby enhancing the user's learning engagement and acceptance.

[0057] This invention obtains the user's post-learning score and answer records for each knowledge node after the user completes learning and answers. It then calculates a learning bias score by combining the two, and processes the score through a second preset model to output recommended learning content. Based on this, a formal learning path is generated, thus constructing a closed-loop adjustment system for learning, feedback, and optimization. This system can iteratively verify and precisely optimize the formal learning path according to the user's actual learning effectiveness and learning bias direction. It overcomes the shortcomings of existing technologies where learning paths are unchanging and lack effectiveness verification, ensuring the continuous relevance and effectiveness of the path content, thereby improving the accuracy of learning content recommendations and the overall learning effect.

[0058] This invention sets the estimated total duration of the formal learning path as the learning termination condition and compares the user's current learning progress in real time. When the learning time is not yet complete, it can proactively calculate the remaining learning time and select supplementary learning content consistent with the determined learning bias from all preset learning content and add it to the formal learning path. This achieves a dynamic and intelligent path adjustment and completion mechanism, which not only avoids the problem of insufficient learning content due to excessively fast learning progress, but also ensures that the supplementary content accurately focuses on the user's weak points or learning bias, thereby maximizing the use of learning time and ensuring the integrity and depth of learning. Attached Figure Description

[0059] Figure 1 This is a flowchart of the method of the present invention;

[0060] Figure 2 This is a system module diagram of the present invention. Detailed Implementation

[0061] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0062] Example 1

[0063] See Figure 1This embodiment provides a method for generating employment skills learning paths based on machine question answering. By analyzing the user's historical answering process and learning status parameters, the method customizes and dynamically adjusts the learning path for the user to improve the user's learning efficiency and depth. The method specifically includes the following steps:

[0064] The system obtains the user's historical answering process, which is a data set that comprehensively records the user's interaction details in the question-and-answer training. It includes the user's answering time, answering order, participation ratio of different question types (such as multiple choice and subjective questions), and specific behaviors such as clicking, browsing, and skipping questions.

[0065] The acquired historical response process is then processed by a set of preset cleaning rules to remove invalid information of preset types, such as data records with incorrect format or invalid interaction records with zero response time, thereby generating a structured data package to be analyzed. This structured data package to be analyzed specifically refers to the data set generated after the historical response process has been processed by the cleaning rules, which has a uniform format and no invalid information, and is used as the direct input for subsequent feature processing.

[0066] Based on a set of preset feature processing rules, namely a set of preset processing logic for extracting and generating answer preference tags from structured data packets to be analyzed, the data packets to be analyzed are processed to generate answer preference tags. Specifically, answer preference tags include answer time tags and answer accuracy tags. Answer preference tags refer to data tags that quantitatively describe user answer preferences. In this scheme, they specifically include answer time tags and answer accuracy tags. The answer time tag refers to a classification tag (e.g., fast, medium, slow) that represents the user's answering pace, generated by comparing the user's average answering time with preset time segmentation standards. The generation process of both is as follows:

[0067] Based on the extracted answer time and the total number of questions in the answer session, the user's average answer time is calculated. This average answer time is then compared with a preset time segmentation standard to generate an answer time tag that represents the user's answering rhythm. The time segmentation standard refers to the preset criteria used to map continuous average answer time values ​​to discrete answer time tags. For example, an average answer time of less than 30 seconds is defined as fast, 30 to 90 seconds as medium, and more than 90 seconds as slow.

[0068] Based on historical answer results and combined with pre-defined question complexity parameters for each question (such as those quantified as the depth of knowledge points, the number of steps required to solve the problem, or the historical average error rate), a weighted calculation is performed to generate an accuracy label. This label is used to accurately represent the accuracy of the user's overall performance. The weighted calculation specifically involves: summing the question complexity parameter values ​​for each question answered correctly to obtain a total score; summing the question complexity parameter values ​​for all answered questions to obtain a theoretical total score; and finally dividing the total correct score by the theoretical total score to obtain an accuracy rate that more accurately represents the user's overall performance on questions of different difficulty levels. The answer time label and the answer accuracy label together constitute a quantitative description of the user's answer preferences, providing a key basis for the generation of subsequent personalized paths.

[0069] After obtaining the user's historical answers, a set of preset scoring rules is used, that is, a set of preset algorithms or logic for calculating the user's answer score based on the user's historical answer results, to calculate the user's answer score based on the historical answer results.

[0070] To more accurately assess users' ability levels, user scores are further refined into a score to be verified and a reference score. Specifically, the score to be verified is the user's average score over a preset time period, which is a direct measure of the user's current ability level; while the reference score is the standard minimum score provided by the system within the same preset time period, which represents the minimum ability baseline required to master the knowledge at the current stage.

[0071] The system employs a pre-defined calculation process to determine the difference between the score to be verified and the reference score. This difference is defined as the user's overall learning score, which visually reflects the gap between the user's current level and the expected level. For example, a positive value indicates that the user's performance exceeds expectations, while a negative value suggests that the user needs to strengthen their learning.

[0072] The user's answer score, which includes the score to be verified, the reference score, and the user's overall learning score, is uniformly integrated and defined as a learning status parameter. This parameter is used to comprehensively describe the user's current learning status and generate a data package to be verified containing the learning status parameter. This data package is used as one of the core inputs in the subsequent path generation process to reflect the user's ability level.

[0073] Taking the obtained answer preference tags and the data package to be verified as input, an initial formal learning path that matches the user's current ability and preferences is generated according to a preset path generation function. The path generation function comprehensively considers the complexity of the questions, the relevance of knowledge points (a preset structured data describing the pre- and post-dependent relationships of knowledge points to ensure the logical coherence of the learning path), answer preference tags, and the data package to be verified. The specific processing procedure is as follows:

[0074] Based on the overall learning score of users in the data package to be verified, knowledge nodes with low user mastery are selected as learning items.

[0075] Based on the pre-defined relevance of knowledge points, the learning items are sorted to ensure that the learning order conforms to the logical sequence of knowledge.

[0076] Based on the answer preference tags, questions of appropriate difficulty are matched for each sorted knowledge node. Specifically, for users with high accuracy and short answer times, questions with higher difficulty parameters are matched, while those with lower accuracy and short answer times are matched. After this initial formal learning path is established, users can learn and answer questions based on it. During the user's learning process, the accuracy rate of their answers at each knowledge node is recorded, and their post-learning score is calculated based on the question weight, question difficulty parameter, and error rate, forming a set of learning path score data. This data set, consisting of the post-learning scores obtained by the user at each knowledge node in the learning path, is used for the next stage of bias analysis.

[0077] The path construction logic is a set of preset rules that comprehensively consider the user's overall learning score, the relevance of knowledge points, and answer preference tags. These rules are used to filter knowledge nodes, sort and match questions, and finally generate a personalized learning path.

[0078] To further refine the recommended content, an in-depth analysis of users' learning preferences is conducted:

[0079] After a user completes the initial formal learning path, the system retrieves the user's interaction records with the content (video explanations, text and image materials) associated with each knowledge node within the initial formal learning path. Based on these interaction records, a learning bias score is calculated. This learning bias score is a numerical value quantifying the user's preference for a particular content format, obtained by calculating the ratio of the user's actual interaction time with a specific content format to the standard interaction time. The calculation process is as follows:

[0080] For each content format, the ratio of the user's actual interaction time to the standard duration of that content is calculated, and this ratio quantifies the user's preference for that content format.

[0081] Based on the acquired post-learning scores, the rating results of each knowledge node in the initial formal learning path are determined. The rating results are used to classify and evaluate the user's mastery of each knowledge node according to the post-learning scores (i.e., nodes with scores above 90 are rated as mastered, nodes with scores between 70 and 90 are rated as familiar, and nodes with scores below 70 are rated as needing improvement). At the same time, the learning bias score calculated above is defined as the corresponding bias parameter, which is used to characterize the user's preference for different content formats.

[0082] A pre-set second model is used to generate a learning bias score by combining the rating results and bias parameters of each knowledge node. This score is a weighted calculation that integrates a basic recommendation value and the matching degree between the content format and user preferences, resulting in a final recommendation score for each piece of learning content to be recommended. The core logic of the second model is to set a basic recommendation value for each piece of learning content to be recommended. This basic recommendation value refers to the initial recommendation score set for each piece of learning content to be recommended in the scoring conversion rules, and its value is mainly determined by the rating result of the corresponding knowledge node. Weighting is given based on whether the content format matches the user's bias parameters. For example, for a knowledge node whose rating result is "needs to be strengthened," its associated review content will receive a higher basic recommendation value. If the format of the content (such as video) happens to be the user's preference, its final learning bias score will be further improved. After summarizing the learning bias scores of all knowledge nodes, a screening and sorting process is executed: all candidate course content is sorted in descending order according to its learning bias score, and the content with the highest score is selected as recommended learning content for the construction of the formal learning path in the next stage.

[0083] The rating conversion rule refers to a set of preset logic that combines the rating results of knowledge nodes and bias parameters to generate learning bias scores. Its core is to set a basic recommendation value for the content and weight it according to the degree of preference matching.

[0084] Based on the recommended learning content generated in the previous stage, a formal learning path is constructed for the user to execute. This process includes:

[0085] By statistically analyzing the quantity or proportion of different categories of content (e.g., theoretical explanations, case studies) in the recommended learning content, the overall bias of the current recommended content can be determined. The bias refers to the overall content type tendency of the current recommended content, which is determined by statistically analyzing the quantity or proportion of different categories of content in the recommended learning content.

[0086] The recommended learning content is combined in a preset order (based on knowledge point dependencies or the order of the standard teaching syllabus) to generate a formal learning path; the formal learning path refers to the learning plan generated based on the recommended learning content and combined in a preset order, which is for the user to finally execute.

[0087] Obtain the estimated total duration of the formal learning path and set this duration as the learning termination condition; where the learning termination condition refers to the time benchmark set based on the estimated total duration of the formal learning path to determine whether the user's learning has been completed.

[0088] During the user's learning process, the current learning progress duration is obtained in real time and compared with the learning termination conditions.

[0089] If the current learning progress duration is greater than or equal to the duration corresponding to the learning termination condition, the formal learning path remains unchanged. Conversely, if the current learning progress duration is less than the duration corresponding to the learning termination condition, the remaining learning time is calculated, and supplementary learning content that is consistent with the determined bias direction and is not yet included in the current path is selected from all preset learning content. This supplementary learning content is added to the end of the formal learning path to ensure full utilization of the learning time. Here, supplementary learning content refers to additional learning materials selected from all learning content that are consistent with the current bias direction and are not included in the formal learning path when the user's learning progress has not reached the learning termination condition.

[0090] To continuously optimize recommendation performance, this method also includes a closed-loop evaluation and adjustment mechanism for recommendation accuracy:

[0091] After each learning cycle, the review content that the user actually selected and completed after receiving the recommended learning content is obtained, and the path composed of the review content is defined as the user's actual review path, which serves as the benchmark path for evaluation.

[0092] By calculating the number of intersections between the recommended learning content and the user's actual selected baseline path, and dividing this intersection by the number of unions of their respective knowledge nodes, a quantitative overlap is obtained. After normalization, this overlap is used to generate a recommendation accuracy score. This score serves as a feedback indicator for dynamically adjusting the subsequent recommendation generation logic. Specifically:

[0093] If the recommendation accuracy score remains below the preset confidence threshold, the system will be triggered to adjust the parameters of the relevant calculation rules. For example, the parameters of the second preset model used to generate the learning bias score can be adjusted (e.g., increasing the weighting value of preference matching), or the screening criteria used to determine the initial formal learning path can be adjusted (e.g., prioritizing content with lower question complexity parameters when the user's level is low). This will achieve iterative optimization of the recommendation generation mechanism to gradually improve the accuracy of the recommendations. The confidence threshold refers to the preset critical value used for comparison with the recommendation accuracy score.

[0094] In summary, this invention is based on the analysis of users' historical answering process and learning preferences. Starting from the analysis of historical answering process, combined with preset scoring rules and preference recognition mechanisms, it reconstructs the user's learning path, which is practically operable and can be widely applied in online skills training, job competency enhancement and corporate training scenarios.

[0095] Example 2

[0096] See Figure 2As shown, this embodiment provides a machine question-and-answer-based job skills learning path generation system. This system can dynamically generate and adjust the formal learning path based on the user's historical answering process and learning status parameters, thereby improving the efficiency and effectiveness of job skills learning. In its implementation, this system can be deployed on cloud servers, personal computers, or mobile smart terminals (such as smartphones and tablets), and integrated into various online education platforms or enterprise internal training systems through application programming interfaces (APIs) or software development kits (SDKs). The system includes the following modules:

[0097] User Status Acquisition Module

[0098] This module is used to obtain the user's historical answering process and historical answering results. Specifically, this module communicates with the front-end application that interacts with the user in machine question-and-answer sessions to collect detailed interaction data of the user in the historical learning or assessment process. The historical answering process includes the start and end time of the user's answer to each question, the interaction behavior (such as whether to view the hints), and the content of the submitted answer. The historical answering results clearly record whether the user's answer to each question is correct. The historical answering process and historical answering results provide the foundation for all subsequent personalized analysis and path generation.

[0099] Learning path generation module

[0100] This module generates an initial formal learning path for the user based on the historical answering process and results provided by the user status acquisition module. Its specific execution process is as follows:

[0101] The historical answering process is processed to generate answer preference tags. In a specific implementation, the user's answering time and historical answering results are extracted from the historical answering process. Based on the answering time, the average answering time of the user is calculated to generate answering time tags (e.g., fast, medium, careful). At the same time, the user's historical answering results are weighted and calculated according to the question complexity parameter of each question to generate answering accuracy tags (e.g., high accuracy on "difficult" questions in the "data structure" category).

[0102] Based on historical responses, the system calculates the user's score and uses it as a learning status parameter. In one optional implementation, the user's score may include a score to be verified (e.g., the score from the user's most recent diagnostic test) and a reference score (e.g., the user's long-term historical average score on the platform). The module uses a preset scoring function to combine the score to be verified and the reference score to calculate a more comprehensive overall learning score for the user. The determination of the learning status parameter is further based on this overall learning score, thereby more accurately reflecting the user's current level of knowledge mastery.

[0103] The generated answer preference tags and the determined learning state parameters are input into a path generation function. This path generation function can be a decision engine based on expert rules or a trained machine learning model. Based on the input parameters, the function selects and sorts a series of learning resources most suitable for the current user from all preset learning content (a knowledge base containing all available courses, articles, videos, and exercises) to generate an initial formal learning path.

[0104] Learning outcome assessment module

[0105] This module is used to evaluate the learning effect of users after they have learned along the formal learning path, in order to generate a learning bias score and a recommendation accuracy score, providing a basis for dynamic adjustment of the path and optimization of the system itself. In terms of generating a learning bias score, after users complete the learning tasks in the formal learning path and answer the relevant questions, the module obtains the user's answer accuracy rate and converts it into a post-learning score. At the same time, it analyzes the user's answer records at each knowledge node in the path and calculates a learning bias score, which can reflect the differences in the user's performance on different types or difficulty levels of knowledge points.

[0106] Combining the post-learning score with the learning bias score, a final learning bias score is generated and output to the path dynamic adjustment module. In generating the recommendation accuracy score, a feedback verification and model optimization process is executed. This process obtains the recommended learning content generated by the system and simultaneously obtains the review content actually selected by the user from the user's subsequent actual operations, which is the user's actual review path. By comparing the recommended learning content with the user's actual review path, the overlap between the two is calculated, and a recommendation accuracy score is generated based on this overlap. This score quantifies the accuracy of the system's recommendations.

[0107] Path dynamic adjustment module

[0108] This module is triggered in response to specific conditions to dynamically update the user's formal learning path. Its workflow is as follows:

[0109] The system monitors the user's current learning progress duration. For each formal learning path, a learning termination condition is set, which can be set based on the estimated total duration of the path. When the user's current learning progress duration is less than the learning termination condition (for example, the user completes the planned learning task ahead of schedule), this module is activated, indicating that the user has additional learning potential and time.

[0110] It receives a learning bias score generated by the learning effectiveness evaluation module, inputs this score into a second preset model (classification or ranking model), and outputs a set of recommended learning content. It analyzes the distribution of this set of recommended learning content among all preset learning content to determine the user's learning bias direction. Specifically, if the recommended content is mostly in-depth content on the current knowledge point, the learning bias direction is "vertical in-depth"; if the recommended content involves new related knowledge areas, the learning bias direction is "horizontal expansion".

[0111] Based on the determined learning bias, corresponding supplementary learning content is selected from all preset learning content. During the selection process, the difference between the current learning progress time and the learning termination condition is calculated to determine the remaining learning time. The selection scale of the supplementary learning content is ensured to be constrained by the remaining learning time to avoid overburdening the user. After selection, the supplementary learning content is added to the user's formal learning path to complete the path update.

[0112] In addition, the system is also configured to adjust the second preset model based on the recommendation accuracy score generated by the learning effect evaluation module. For example, if the recommendation accuracy score is low, the system can automatically trigger fine-tuning of the parameters of the second preset model or retrain it using new user feedback data, thereby continuously improving the quality of recommendations.

[0113] Through the collaborative work of the aforementioned user status acquisition module, learning path generation module, learning effect evaluation module, and path dynamic adjustment module, this invention can construct a closed-loop learning support system from initial path generation to dynamic adjustment, and then to effect evaluation and model optimization. It can not only customize a personalized starting point based on the user's historical answering process and answer preference tags, but also intelligently respond to the user's current learning progress duration and learning bias score during the learning process, dynamically supplementing learning content, thereby maximizing the use of learning time and improving the depth and breadth of skill mastery. It is suitable for online vocational education and corporate training scenarios that require continuous skill improvement and personalized tutoring.

[0114] Example 3

[0115] This embodiment provides a terminal for generating employment skills learning paths based on machine question answering, including a processor and a memory communicatively connected to the processor. The memory stores a computer program, and when the computer program is executed by the processor, it implements a method for generating employment skills learning paths based on machine question answering.

[0116] The above description is merely a preferred embodiment of this application and is not intended to limit this application. For those skilled in the art, this application can have various modifications and variations. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A method for generating employment skills learning paths based on machine question answering, characterized in that, Includes the following steps: Based on the user's historical answer process and results, an initial formal learning path is generated; And when the user's current learning progress time is detected to be less than the learning termination condition set for the formal learning path, dynamic adjustment of the learning path is executed, which includes: Based on the user's answer records in the formal learning path, determine the learning bias direction; Based on the learning preferences, supplementary learning content is selected from all the preset learning content to update the formal learning path; Among these, determining the learning bias direction based on the user's answer records in the formal learning path includes: After the user completes the learning and answers questions according to the formal learning path, the user's answer accuracy rate is obtained to generate a post-learning score; The learning bias score is calculated based on the user's answer records for each knowledge node completed in the formal learning path. A learning bias score is generated by combining the learning bias score with the score after learning. The learning bias score is input into the second preset model to output recommended learning content; Based on all the preset learning content, the distribution of recommended learning content within the preset learning content is analyzed to determine the direction of learning bias.

2. The method for generating employment skills learning paths based on machine question answering according to claim 1, characterized in that, Based on the user's historical answering process and results, the initial formal learning path is generated, including: Based on answer preference tags and learning state parameters, an initial formal learning path is generated through a path generation function.

3. The method for generating employment skills learning paths based on machine question answering according to claim 2, characterized in that, The historical response process is processed to generate response preference tags; Based on historical responses, user scores are calculated as learning status parameters.

4. The method for generating employment skills learning paths based on machine question answering according to claim 3, characterized in that, Answer preference tags include answer time tags and answer accuracy tags; The process of processing historical responses to generate answer preference tags includes: Extract the answering time and answering result from the historical answering process; Calculate the average answering time based on the answering time to generate answering time tags; The accuracy rate label is generated by weighting the answers with the pre-set complexity of the corresponding test questions.

5. The method for generating employment skills learning paths based on machine question answering according to claim 3, characterized in that, The user's answer score includes a score to be verified and a reference score; Also includes: The user's overall learning score is calculated by combining the score to be verified with the reference score using a scoring function. The determination of learning status parameters is further based on the user's overall learning score.

6. The method for generating employment skills learning paths based on machine question answering according to claim 1, characterized in that, The learning termination criteria are set based on the estimated total duration of the formal learning path; Based on learning preferences, supplementary learning content is selected from the pre-set learning materials to update the formal learning path, including: Calculate the current learning progress duration and the learning termination conditions to determine the remaining learning time; Supplementary learning content is added to the formal learning path, where the selection of supplementary learning content is constrained by the remaining learning time.

7. The method for generating employment skills learning paths based on machine question answering according to claim 1, characterized in that, The method further includes: Obtain the actual review content selected by the user after receiving the recommended learning content, as the user's actual review path; The recommended learning content is compared with the user's actual review path to calculate the overlap between the two contents, and a recommendation accuracy score is generated based on the overlap. The second preset model is adjusted based on the recommendation accuracy score.

8. A machine question-answering-based employment skills learning path generation system, characterized in that, Includes the following modules: The user status acquisition module is used to acquire the user's historical answering process and historical answering results; The learning path generation module is used to generate an initial formal learning path based on the historical answering process and historical answering results; The learning effectiveness evaluation module is used to evaluate the learning effectiveness after the user learns along the formal learning path to generate a learning bias score and a recommendation accuracy score. The path dynamic adjustment module, in response to the detection that the user's current learning progress time is less than the learning termination condition set for the formal learning path, is configured as follows: Based on the learning bias score generated by the learning effectiveness evaluation module, the direction of learning bias is determined; Supplementary learning content is selected based on learning preferences to update the formal learning path; The system is also configured to adjust the second preset model based on the recommendation accuracy score generated by the learning effect evaluation module.

9. The employment skills learning path generation system based on machine question answering according to claim 8, characterized in that, Based on the historical response process and results, the initial formal learning path is generated as follows: Process historical responses to generate answer preference tags; Based on historical responses, calculate the user's response score as a learning status parameter; Based on answer preference tags and learning status parameters, an initial formal learning path is generated; The assessment of learning outcomes to generate learning bias scores and recommendation accuracy scores includes: Obtain the accuracy rate of user responses to generate a post-learning score; Based on the user's answer records for each knowledge node completed in the formal learning path, a learning bias score is calculated, and then the post-learning score is combined with the learning bias score to generate a learning bias rating. Obtain the user's actual review path and compare it with the recommended learning content to generate a recommendation accuracy score.

10. A terminal for generating employment skills learning paths based on machine question answering, characterized in that: It includes a processor and a memory communicatively connected to the processor, the memory storing a computer program, which, when executed by the processor, implements the machine question-answering-based employment skills learning path generation method as described in any one of claims 1 to 7.

Images