Intelligent training method and device based on adaptive learning, equipment and medium

Abstract

This application discloses an intelligent training method, apparatus, device, and medium based on adaptive learning. The method includes: entering a pre-study phase, collecting user pre-study behavior data on pre-study materials, updating the user's basic ability profile to obtain a first ability profile; entering a practice phase, generating personalized practice tasks based on the first ability profile; collecting the user's first operation data, and updating the first ability profile based on the first operation data to obtain a second ability profile; entering a simulation phase, generating simulated training tasks based on the second ability profile and a preset customer profile; collecting the user's second operation data, and updating the second ability profile based on the second operation data to obtain a third ability profile; entering an examination phase, generating standardized examination tasks based on the third ability profile; collecting the user's third operation data, and generating a user ability assessment report. This application embodiment enables dynamic matching of task content with learner abilities.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, and in particular to intelligent training methods, devices, equipment and media based on adaptive learning. Background Technology

[0002] As companies demand higher efficiency and effectiveness in employee training, traditional training models based primarily on lectures and paper-based tests are no longer sufficient to meet the needs for personalized, high-frequency, and quantifiable assessments. In recent years, AI-driven intelligent learning systems have gradually emerged, with common forms including intelligent question-answering robots, knowledge point recommendation engines, and voice recognition scoring systems.

[0003] Some systems have attempted to use AI for conversation simulation training, but these are mostly limited to single-stage "question-and-answer practice" or "voice scoring." Trainees enter the practice without fully understanding the knowledge points, resulting in low training efficiency. Furthermore, since all trainees face the same question bank and standard answers, the system cannot dynamically adjust the training content based on the trainees' performance, making it difficult to achieve individualized instruction. Summary of the Invention

[0004] This application provides intelligent training methods, devices, equipment, and media based on adaptive learning, aiming to solve the technical problems in the prior art where students directly enter practice without fully understanding the knowledge points, resulting in low training efficiency, and the inability to dynamically adjust training content according to students' performance, making it difficult to achieve individualized instruction.

[0005] In a first aspect, embodiments of this application provide an intelligent training method based on adaptive learning, which includes: In response to a user-triggered pre-study instruction, the system enters the pre-study phase, collects the user's pre-study behavior data for the pre-study materials, and updates the user's basic ability profile based on the pre-study behavior data to obtain a first ability profile. In response to the user-triggered practice instruction, the user enters the practice phase and generates a personalized practice task based on the first ability profile. The personalized practice task includes multiple single-round question-and-answer tasks. Collect the user's first operation data for the personalized practice task, and update the first ability profile based on the first operation data to obtain the second ability profile; In response to the user-triggered simulation command, the simulation phase is entered, and a simulation training task is generated based on the second capability profile and the preset customer profile. The simulation training task is a multi-turn dialogue task. Collect the user's second operation data for the simulated training task, and update the second capability profile based on the second operation data to obtain a third capability profile; In response to the examination instruction triggered by the user, the examination phase is entered, and a standardized examination task is generated based on the third competency profile and the current training objectives. Collect the user's third operational data for the standardized test task, and generate the user's competency assessment report based on the third operational data.

[0006] In some embodiments, before collecting the user's pre-study behavior data regarding the pre-study materials, the method further includes: Obtain the unstructured training materials corresponding to the current coaching task; Natural language processing technology is used to extract key entities, entity relationships, and business rules from the unstructured training materials to construct a domain knowledge graph; The graph centrality algorithm is used to classify the importance of knowledge nodes in the domain knowledge graph and sort them by cognitive dependence to form a knowledge skeleton. By calling a preset standardized teaching template, the knowledge point information in the knowledge skeleton is filled into the corresponding slots of the template to obtain the first initial material; The first initial material is transcribed into a teaching format using a large model to obtain the second initial material; The second initial material is subjected to multimodal structured assembly to generate the pre-study material, which includes at least one of knowledge point cards and flowcharts.

[0007] In some embodiments, the pre-study behavior data includes pre-study completion rate, saved content, question markers, and note content; updating the user's basic ability profile based on the pre-study behavior data to obtain a first ability profile includes: Based on the pre-study completion rate, the saved content, the question marks, and the notes, the user's weaknesses regarding the pre-study materials are determined. The basic capability profile is updated based on the weaknesses to obtain the first capability profile.

[0008] In some embodiments, the first operation data includes the answer content corresponding to each single-round question-and-answer task, the user's emotional state, and the user's body language; updating the first ability profile based on the first operation data to obtain a second ability profile includes: Based on the content of the answer, the user's emotional state, and the user's body language, a multi-dimensional answer score is determined, which assesses the completeness of knowledge points, the logic of the speech, the emotional expressiveness, and the expressiveness of the actions. The first ability profile is updated based on the multi-dimensional response score to obtain the second ability profile.

[0009] In some embodiments, after determining a multi-dimensional response score based on the response content, the user's emotional state, and the user's body language regarding the completeness of knowledge points, the logicality of the speech, the emotional expressiveness, and the expressiveness of actions, the method further includes: Based on the multidimensional response score, response suggestions are generated.

[0010] In some embodiments, generating a simulation training task based on the second capability profile and a preset customer profile includes: Based on the second capability profile, a target simulation scenario is selected from a preset scenario library; The simulation training task is generated based on the target simulation scenario, the preset user profile, and the preset random events.

[0011] In some embodiments, after collecting the user's third operational data for the standardized test task, the method further includes: The user is subjected to cheating detection based on the third operation data, and the cheating detection includes voiceprint recognition detection and screen monitoring detection.

[0012] Secondly, embodiments of this application also provide an intelligent training device based on adaptive learning, comprising: The transceiver unit is used to receive pre-study instructions; The processing unit is configured to: enter a pre-study phase in response to a user-triggered pre-study instruction; collect user pre-study behavior data related to pre-study materials via the transceiver unit and update the user's basic ability profile based on the pre-study behavior data to obtain a first ability profile; enter a practice phase in response to a user-triggered practice instruction; generate personalized practice tasks based on the first ability profile, the personalized practice tasks including multiple single-round question-and-answer tasks; collect first operation data of the user on the personalized practice tasks via the transceiver unit and update the first ability profile based on the first operation data to obtain a second ability profile; enter a simulation phase in response to a user-triggered simulation instruction; generate a simulation training task based on the second ability profile and a preset customer profile, the simulation training task being a multi-round dialogue task; collect second operation data of the user on the simulation training task via the transceiver unit and update the second ability profile based on the second operation data to obtain a third ability profile; enter an examination phase in response to a user-triggered examination instruction; generate a standardized examination task based on the third ability profile and the current training objective; collect third operation data of the user on the standardized examination task via the transceiver unit and generate the user's ability assessment report based on the third operation data.

[0013] Thirdly, embodiments of this application also provide a computer device, which includes a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the above-described method.

[0014] Fourthly, embodiments of this application also provide a computer-readable storage medium storing a computer program, the computer program including program instructions that, when executed by a processor, can implement the above-described method.

[0015] This application provides an intelligent training method, apparatus, device, and medium based on adaptive learning. The method includes: entering a pre-study phase in response to a user-triggered pre-study instruction; collecting the user's pre-study behavior data on pre-study materials and updating the user's basic ability profile based on the pre-study behavior data to obtain a first ability profile; entering a practice phase in response to a user-triggered practice instruction; generating a personalized practice task based on the first ability profile, the personalized practice task including multiple single-round question-and-answer tasks; collecting the user's first operation data on the personalized practice task and updating the first ability profile based on the first operation data to obtain a second ability profile; entering a simulation phase in response to a user-triggered simulation instruction; generating a simulation training task based on the second ability profile and a preset customer profile, the simulation training task being a multi-round dialogue task; collecting the user's second operation data on the simulation training task and updating the second ability profile based on the second operation data to obtain a third ability profile; entering an examination phase in response to a user-triggered examination instruction; generating a standardized examination task based on the third ability profile and the current training objective; collecting the user's third operation data on the standardized examination task and generating a user ability assessment report based on the third operation data. This application embodiment constructs a four-stage progressive learning process, realizing a complete closed loop from knowledge preview to single-point practice, comprehensive simulation, and standardized assessment. Users can improve training efficiency by previewing before entering practice. Furthermore, this embodiment achieves dynamic matching between task content and learner's ability through iterative updates of ability profiles at each stage, truly achieving adaptive learning and enabling individualized instruction. Attached Figure Description

[0016] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1A schematic diagram illustrating an application scenario of the intelligent training method based on adaptive learning provided in this application embodiment; Figure 2 A flowchart illustrating the intelligent training method based on adaptive learning provided in this application embodiment; Figure 3 A schematic diagram of a sub-process of the intelligent training method based on adaptive learning provided in the embodiments of this application; Figure 4 Another sub-process diagram of the intelligent training method based on adaptive learning provided in the embodiments of this application; Figure 5 Another sub-process diagram of the intelligent training method based on adaptive learning provided in the embodiments of this application; Figure 6 A schematic block diagram of an intelligent training device based on adaptive learning provided in an embodiment of this application; Figure 7 A schematic block diagram of a computer device provided in an embodiment of this application. Detailed Implementation

[0018] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this application. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0019] It should be noted that any AI models, software tools, or components not belonging to this company appearing in the embodiments of this application are merely illustrative examples and do not represent actual use. All user personal information involved in the embodiments of this application has been obtained by authorized entities (who have known and consented) or fully authorized by all parties through various legal and compliant means. The collection, storage, use, processing, transmission, provision, and disclosure of the information, data, and signals involved all comply with relevant laws and regulations and do not violate public order and good morals.

[0020] It should be understood that, when used in this specification and the appended claims, the terms "comprising" and "including" indicate the presence of the described features, integrals, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof.

[0021] It should also be understood that the terminology used in this application specification is for the purpose of describing particular embodiments only and is not intended to limit the application. As used in this application specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0022] It should also be further understood that the term “and / or” as used in this application specification and the appended claims means any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0023] This application provides intelligent training methods, apparatus, devices, and media based on adaptive learning.

[0024] The execution entity of the adaptive learning-based intelligent training method can be the adaptive learning-based intelligent training device provided in the embodiments of this application, or a computer device that integrates the adaptive learning-based intelligent training device. The adaptive learning-based intelligent training device can be implemented in hardware or software, and the computer device is the physical device corresponding to the adaptive learning-based intelligent training system.

[0025] The intelligent training method based on adaptive learning provided in this application can be applied to scenarios such as enterprise employee skills training, product knowledge learning, and customer service script practice.

[0026] Please see Figure 1 , Figure 1 This diagram illustrates an application scenario of the adaptive learning-based intelligent training method provided in this embodiment. The adaptive learning-based intelligent training method is applied to... Figure 1In the adaptive learning-based intelligent training system, the system (hereinafter referred to as the system) includes a server 10 and a user terminal 20. The server 10 is used to enter the pre-study phase in response to a user-triggered pre-study instruction, collect the user's pre-study behavior data for the pre-study materials, and update the user's basic ability profile based on the pre-study behavior data to obtain a first ability profile; in response to a user-triggered practice instruction, it enters the practice phase, generates personalized practice tasks based on the first ability profile, the personalized practice tasks including multiple single-round question-and-answer tasks; collects the user's first operation data for the personalized practice tasks, and updates the system based on the first operation data. The first competency profile is used to obtain a second competency profile; in response to a simulation command triggered by the user, a simulation phase is entered, and a simulation training task is generated based on the second competency profile and a preset customer profile. The simulation training task is a multi-turn dialogue task; second operation data of the user on the simulation training task is collected, and the second competency profile is updated based on the second operation data to obtain a third competency profile; in response to an examination command triggered by the user, an examination phase is entered, and a standardized examination task is generated based on the third competency profile and the current training objective; third operation data of the user on the standardized examination task is collected, and a competency assessment report of the user is generated based on the third operation data.

[0027] The user terminal 20 is used to switch between practice stages by triggering pre-study instructions, practice instructions, simulation instructions, and exam instructions. At different stages, it obtains personalized practice tasks, personalized practice tasks, simulation training tasks, and standardized exam tasks issued by the server 10, and obtains the user's operation data with each task, such as pre-study behavior data, first operation data, second operation data, and third operation data, and feeds the obtained operation data back to the server 10.

[0028] In this embodiment, the system adopts a front-end and back-end separated microservice architecture, including a front-end application layer (located in user terminal 20), an AI engine layer (located in server 10), and a data storage layer (located in server 10 or other storage devices). The data storage layer includes: relational databases (for storing user information and training records), graph databases (for storing knowledge graphs), and object storage (for storing audio and video files). The AI ​​engine layer includes a knowledge graph construction module, a pre-learning content generation module, a dynamic question bank generation module, a simulation scenario generation engine, a multimodal evaluation model, a competency profile model, and a path recommendation engine.

[0029] The user terminal 20 provides a student interaction interface that supports functions such as video playback, voice input, text input, and real-time feedback display. It can be used with mobile phones, computers, or AR / VR devices, and can provide a more immersive training experience through AR / VR devices.

[0030] Figure 2 This is a flowchart illustrating the intelligent training method based on adaptive learning provided in an embodiment of this application. Figure 2 As shown, the method includes the following steps S110-S170.

[0031] S110. In response to the user-triggered pre-study instruction, enter the pre-study stage, collect the user's pre-study behavior data for the pre-study materials, and update the user's basic ability profile based on the pre-study behavior data to obtain a first ability profile.

[0032] In this embodiment, the system can enter the pre-study stage based on the pre-study instruction triggered by the user. The pre-study instruction contains the task identifier of the current training task. Then, the system determines the pre-study material corresponding to the task identifier from the server's database and sends the pre-study material to the user's corresponding user terminal. The user then displays the pre-study material on the user terminal and performs pre-study. The system collects the user's pre-study behavior data in real time through the user terminal. The pre-study behavior data includes at least the pre-study completion rate, saved content, question marks, and notes. The system updates the user's basic ability profile in real time based on the user's pre-study behavior data.

[0033] Among them, the basic ability profile is an initial ability profile pre-built by the system for users, which is used to represent the initial profile of users before they undertake the current training task and serves as the basis for subsequent adaptive training; the pre-learning stage is used for users to complete knowledge preview and understand basic knowledge points before formal practice, in order to prepare for subsequent skills training.

[0034] In one specific embodiment, the system provides a pre-study entry point on the front-end interface of the user terminal. The user triggers the pre-study instruction by clicking, touching, or confirming. The user terminal reports the pre-study instruction to the server. After the server parses the instruction, it retrieves the structured pre-study materials corresponding to the current training task and pushes them to the user's front end for display, and the system officially enters the pre-study stage.

[0035] During the pre-study process, the system collects user pre-study behavior data in real time and incrementally updates the user's initial basic ability profile based on the collected pre-study behavior data. The specific updates include: Identify weak knowledge points based on question marks and notes; The level of engagement in learning can be judged by the completion rate of pre-study; Based on the content collected, determine the key areas of focus; The system writes the above analysis results into the ability profile, forming a first ability profile that is more in line with the user's actual knowledge level.

[0036] Among them, the first capability profile mainly represents the user's level of knowledge mastery.

[0037] In some embodiments, before collecting the user's pre-study behavior data regarding the pre-study materials, the method further includes: The process involves: acquiring unstructured training materials corresponding to the current training task; extracting key entities, entity relationships, and business rules from the unstructured training materials using natural language processing technology to construct a domain knowledge graph; ranking the knowledge point nodes in the domain knowledge graph by importance and cognitive dependency using a graph centrality algorithm to form a knowledge skeleton; calling a preset standardized teaching template to fill the knowledge point information in the knowledge skeleton into the corresponding slots of the template to obtain the first initial material; transcribing the first initial material into a teaching format using a large model to obtain the second initial material; and performing multimodal structured assembly on the second initial material to generate the pre-study material, which includes at least one of knowledge point cards and flowcharts.

[0038] Specifically, in generating pre-study materials, this embodiment first acquires unstructured training materials (PDF, PPT, Word, etc.); extracts key entities, relationships, and business rules using natural language processing technology to construct a domain knowledge graph; uses a graph centrality algorithm to classify the importance of knowledge point nodes and rank their cognitive dependencies to form a knowledge skeleton; calls standardized teaching templates and fills the knowledge point information into the corresponding slots; performs pedagogical transcription through a large model to popularize professional content and refine it with contextualization; and then performs multimodal structured assembly to generate pre-study materials in the form of knowledge point cards, flowcharts, etc.

[0039] The knowledge skeleton, based on a domain knowledge graph, uses a graph centrality algorithm to classify knowledge point nodes by importance and cognitive dependence, resulting in a hierarchical, logically progressive knowledge point structure that conforms to human learning patterns. The standardized teaching template is a pre-defined, fixed-structure framework for arranging teaching content, including predefined slots for titles, key points, steps, examples, and precautions. Multimodal structured assembly combines text content, process steps, key point lists, and diagrams, after being transcribed and validated for instructional purposes, into visually appealing, clearly structured, and directly displayable pre-learning materials according to front-end rendering requirements.

[0040] Furthermore, after the first initial material is transcribed into a teaching format using a large model, in order to ensure the accuracy of the transcribed content, this embodiment also needs to perform a knowledge consistency check on the second initial material obtained after transcription to filter out illusions and conflicting content. Specifically, the knowledge consistency check refers to comparing and verifying the content transcribed into a teaching format using the large model with the constructed domain knowledge graph and the original training materials, removing AI illusion information that does not conform to the original knowledge, and eliminating problems such as inconsistencies, logical conflicts, and rule inconsistencies in the content, so as to ensure that the final generated pre-study materials are authentic, accurate, compliant, and consistent with the official knowledge of the enterprise.

[0041] As can be seen, this embodiment automatically generates pre-learning materials, which can directly process unstructured enterprise data, reducing enterprise usage costs and configuration thresholds; based on knowledge graphs and graph ranking algorithms, it automatically sorts out learning logic and key points, forming learning content that conforms to cognitive laws; through template filling, large model polishing and multimodal assembly, it automatically generates structured, easy-to-understand, and visual pre-learning materials, eliminating the need for manual courseware production; and the entire process is automated, improving the efficiency of training content construction and ensuring content standardization and consistency.

[0042] Furthermore, in some embodiments, please refer to Figure 3 The first ability profile is obtained by updating it through the following steps: S1101. Determine the user's weaknesses regarding the pre-study materials based on the pre-study completion rate, the saved content, the question marks, and the notes content. S1102. Update the basic capability profile based on the weaknesses to obtain the first capability profile.

[0043] In this embodiment, the pre-study behavior data includes pre-study completion rate, saved content, question marks, and notes content; the system identifies the student's weak knowledge points based on the above data, and updates the basic ability profile based on the weak points to obtain the first ability profile.

[0044] In one embodiment, specifically, the system uses pre-study completion statistics to determine the overall browsing rate of the current pre-study material, the time spent on each knowledge point, and whether all content was viewed completely. If the completion rate is low or the time spent on key knowledge points is too short, it indicates that the user has not fully studied or understood this part of the knowledge, and it is marked as a weakness. Saved content refers to knowledge points actively saved by the user, representing content that the user considers important, difficult to understand, and requires focused memorization. The system directly marks saved knowledge points as key areas requiring intensive study and includes them in the scope of weaknesses. Question marks added by the user to specific knowledge points during pre-study directly reflect content that the user is confused about, does not understand clearly, or cannot understand independently. The knowledge points corresponding to these question marks can be directly identified as core weaknesses. Notes include text or voice notes recorded by the user. The system extracts keywords using natural language processing technology and compares them with standard knowledge points in the knowledge graph. If the notes are incomplete, contain errors, or omit key information, it is determined that the knowledge point is not accurately grasped and is listed as a weakness. The system integrates and analyzes the above four types of data to form a personalized list of weaknesses for the user at this stage, reflecting the knowledge points not mastered, areas of misunderstanding, and content that needs to be strengthened in the first ability profile.

[0045] This embodiment can mine weaknesses from multi-dimensional behavioral data, making the resulting first ability profile more accurate; and by completing the weakness location in the pre-learning stage, subsequent practice tasks can be directly targeted to strengthen the weaknesses, thereby improving training efficiency.

[0046] In some embodiments, the generation logic of pre-study materials can be optimized by accumulating student notes and question data, forming a closed-loop iteration.

[0047] S120. In response to the practice instruction triggered by the user, enter the practice stage and generate a personalized practice task based on the first ability profile. The personalized practice task includes multiple single-round question and answer tasks.

[0048] In this embodiment, after the user completes the preview, the user triggers the practice instruction through the front-end operation. After receiving the instruction, the server ends the preview stage, switches to the practice stage, and loads the practice interactive interface. The practice stage is used for single-point, repetitive, and intensive training on a single knowledge point.

[0049] Furthermore, the system reads the first ability profile and dynamically generates personalized practice tasks based on the recorded weaknesses, mastery level, and knowledge level. In these personalized practice tasks, questions related to weak knowledge points are prioritized, the difficulty of the questions is matched with the user's ability, and the task format is a single-round question and answer, i.e., one question and one answer, one feedback, and one-point reinforcement. The practice tasks are used to specifically strengthen the knowledge gaps identified in the pre-study stage.

[0050] Specifically, the system generates personalized practice tasks based on the student's pre-study completion rate, note content, and historical incorrect questions (incorrect questions from previous training sessions), using a path recommendation engine. The dynamic question bank generation module selects or combines questions from the question bank to generate questions (such as "Please introduce product A using three selling points").

[0051] S130. Collect the user's first operation data for the personalized practice task, and update the first ability profile based on the first operation data to obtain the second ability profile.

[0052] During the practice, the system collects the user's first operation data, including the answers corresponding to each single round of question and answer tasks, the user's emotional state, and the user's body language. Then, based on the collected first operation data, the system updates the first ability profile to obtain the second ability profile.

[0053] Specifically, trainees (users) answer questions via voice or text on their user terminals, upload their answers to the server, and upload video of the user answering questions to the server via the user terminal's visual module.

[0054] In some embodiments, specifically, please refer to Figure 4 The second capability profile is obtained through the following steps: S1301. Determine a multi-dimensional answer score based on the answer content, the user's emotional state, and the user's body language, considering the completeness of knowledge points, the logic of the speech, the emotional expressiveness, and the expressiveness of actions. S1302. Update the first ability profile based on the multi-dimensional answer score to obtain the second ability profile.

[0055] Specifically, the multimodal evaluation model in the system constructs a multi-task learning framework that integrates speech, vision, and text. Through ASR speech recognition, computer vision facial expression / motion capture, and emotion computing, it integrates multi-dimensional features such as student answers, emotional state, learning attitude, and body language. Based on the attention mechanism, it accurately identifies key information and outputs fine-grained scores in multiple dimensions, such as knowledge point completeness, speech logic, emotional expressiveness, and motor expressiveness. Based on these scores, the primary ability profile is updated, and targeted improvement suggestions are generated and recorded in the student's file simultaneously (e.g., completeness of selling points: 80%; speech logic: 70%; tone of friendliness: 60%).

[0056] The system incrementally updates the first ability profile, supplementing it with dimensions such as skill proficiency, communication ability, and adaptability to form a second ability profile. The second ability profile mainly represents the user's mastery level of a single skill.

[0057] This embodiment breaks through the traditional model of scoring only based on the correctness of the answer, and realizes a multimodal and detailed comprehensive evaluation; and integrates language, emotion and body language evaluation, which is closer to the real job service scenario.

[0058] Specifically, after the update, the first ability profile is officially upgraded to the second ability profile. Its core feature is that the first ability profile, which "only represents the degree of knowledge mastery", is upgraded to the second ability profile, which "represents the user's mastery level of a single skill". It retains the basic knowledge information in the pre-study stage and supplements the skill performance data in the practice stage, forming a traceable and iterative skill ability file, which serves as the core basis for entering the simulation stage and realizes the synchronous iteration of the ability profile and the user's training progress.

[0059] In one embodiment, this embodiment employs a multimodal evaluation model to determine multidimensional response scores for knowledge point completeness, verbal logic, emotional expressiveness, and motor expressiveness. This multimodal evaluation model uses a multi-task learning framework to simultaneously predict multiple scoring dimensions: Specifically, the evaluation model is built on a multi-task learning (MTL) framework, using a shared feature encoder as the basic module and connecting to multiple independent scoring heads.

[0060] In a forward inference process, multiple dimensions such as the completeness of knowledge points, the logic of the speech, the emotional expressiveness, and the expressiveness of actions can be quantitatively scored simultaneously. This not only improves the evaluation efficiency, but also enhances the generalization ability of the model through feature sharing between tasks, avoiding the one-sidedness of single-dimensional evaluation.

[0061] Furthermore, this embodiment uses an attention mechanism to identify missing key information in answers when scoring the completeness of knowledge points: Specifically, through multi-head attention, the model can dynamically focus on the alignment between the answer and standard knowledge points: on the one hand, it locates the core semantic units in the answer; on the other hand, it compares them with a pre-set key knowledge point graph to accurately identify missing key selling points, necessary process steps, compliant expressions, and other key information, providing a precise basis for generating targeted completion suggestions in the future. In some embodiments, after multidimensional response scoring, the method further includes: generating response suggestions based on the multidimensional response score.

[0062] For example, it can be suggested to add the selling point of "energy saving". Users can retrain the current single-round question and answer task based on the answer suggestion, and the suggestion and the corresponding training information can be recorded in the student's file.

[0063] This embodiment enables real-time error correction and guidance, allowing trainees to quickly identify problems and make improvements, thus enhancing the effectiveness of practice.

[0064] S140. In response to the simulation command triggered by the user, enter the simulation phase and generate a simulation training task based on the second capability profile and the preset customer profile. The simulation training task is a multi-turn dialogue task.

[0065] In this embodiment, after the user completes the exercise, a simulation command is triggered. After receiving the command, the server enters a highly realistic combat simulation stage, loading a multi-round dialogue interaction interface. The simulation stage is used for comprehensive skill application, scenario-based response, and process-oriented training.

[0066] In some embodiments, please refer to Figure 5 The simulation training task is generated through the following steps: S1401. Select a target simulation scenario from the preset scenario library based on the second capability profile; S1402. Generate the simulation training task based on the target simulation scenario, the preset user profile, and the preset random events.

[0067] Specifically, the system matches simulated scenarios of corresponding difficulty based on the weak skills identified in the second competency profile. For example, based on the trainee's weaknesses during the practice phase (such as a weak ability to handle price objections), the system selects or generates targeted simulated scenarios from the scenario library (e.g., "How to respond when a customer says a competitor's product is cheaper?"). It also combines preset customer profiles (personality, budget, usage habits) with random events (such as "receiving a sudden phone call") to generate dynamic, multi-round, and evolving simulated training tasks. These simulated training tasks involve multi-round dialogues, requiring users to continuously respond, advance the process, and handle customer issues, thus more closely resembling real-world work scenarios.

[0068] During the simulation phase, users can engage in multiple rounds of voice dialogue with the AI ​​client, and the system recognizes intent and advances the plot in real time.

[0069] This embodiment can match corresponding simulation scenarios to trainees' skill gaps to achieve targeted simulation training; and the introduction of random events and dynamic customer profiles enhances the realism of the scenarios and the effectiveness of response training, making the training closer to actual work.

[0070] In some embodiments, after the dialogue ends, the system also performs a comprehensive evaluation based on the complete dialogue log and generates a "simulated combat report".

[0071] S150. Collect the user's second operation data for the simulated training task, and update the second capability profile based on the second operation data to obtain a third capability profile.

[0072] During the multi-round dialogue, the system collects second operational data, including: the complete content of the multi-round dialogue, process compliance, accuracy of the wording, emotional response ability, performance in handling unexpected problems, and dialogue completion rate.

[0073] The second set of operational data is used to evaluate the user's overall practical skills.

[0074] Then, based on the overall performance during the simulation phase, the system updates the second capability profile. For example, it adds practical ability scores, process compliance scores, and comprehensive adaptability scores to the second capability profile, and marks any remaining weaknesses. Finally, a third capability profile is formed, which represents the user's complete job competency.

[0075] S160. In response to the examination instruction triggered by the user, enter the examination stage and generate a standardized examination task based on the third competency profile and the current training objectives.

[0076] After a user triggers the exam command, the system enters a standardized and formal competency certification phase. This exam phase is independent of the training phase and is used to objectively evaluate the final learning outcome, achieving separation of training and testing. The system combines a third competency profile (user's competency level) and job training objectives (essential knowledge points and skill requirements) to generate comprehensive, appropriately challenging, and standardized exam tasks, including objective questions and subjective questions via voice / text.

[0077] For example, while ensuring comprehensive coverage of test question types, add test questions that address users' weaknesses.

[0078] S170. Collect the user's third operation data for the standardized test task, and generate the user's ability assessment report based on the third operation data.

[0079] In this embodiment, during the test, the system collects the user's third-party operation data, including: answer content, answer duration, operation behavior, and abnormal operations (screen switching, copying, searching for questions, etc.).

[0080] In this embodiment, the system generates standardized test questions (including objective and subjective voice questions) based on training objectives and trainees' historical performance; trainees complete the test within a specified time, and the answers (third operation data) are encrypted and uploaded; the system automatically grades the papers, and subjective questions are scored by a multimodal assessment model; a final competency assessment report is generated, which includes scores for each dimension and a matching degree analysis with the job competency model.

[0081] In some embodiments, after collecting the user's third operational data for the standardized test task, the method further includes: performing cheating detection on the user based on the third operational data, wherein the cheating detection includes voiceprint recognition detection and screen monitoring detection.

[0082] As can be seen, this embodiment, through dual detection of biometrics and behavioral monitoring, can ensure the fairness and rigor of the testing process; and make the results of the competency assessment report reliable, which can be directly used as an objective basis for enterprise recruitment and promotion.

[0083] Furthermore, this application provides a data closed-loop and dynamic adaptation mechanism, including: Full-link behavioral data collection and intelligent analysis: The system continuously collects full-dimensional behavioral data of students in the four stages of pre-study, practice, simulated dialogue and test. Through multimodal feature extraction and behavioral modeling, it completes multi-dimensional intelligent analysis of learning engagement, knowledge mastery, interaction proficiency and test-taking ability, providing full data support for operational decisions.

[0084] Dynamic competency profile iteration: The competency profile model is updated incrementally every 24 hours. Based on the four-stage analysis results, it accurately identifies the individual competency gaps of trainees (such as product knowledge, on-the-spot adaptability, and standardization of communication skills), forming a quantifiable and traceable trainee competency profile.

[0085] Personalized training path dynamic adaptation: The path recommendation engine dynamically adjusts subsequent training plans and resource allocation based on the latest learner ability profiles, and pushes highly adaptable training content to weak abilities (such as adding high-difficulty simulation scenarios for learners with weak adaptability), so as to achieve personalized training for each individual.

[0086] Team-level operational decision support: Enterprise administrators can intuitively grasp the common shortcomings of the team's capabilities, the distribution of capabilities at different levels, and the trend of training effectiveness through the overall team capability heat map. Based on the four-stage aggregated analysis data, they can make operational decisions such as optimizing training content, allocating resources, and providing targeted supplementary training, thereby achieving closed-loop iteration and continuous optimization of the training system.

[0087] In summary, this embodiment of the application constructs a four-stage progressive learning process, realizing a complete closed loop from knowledge preview to single-point practice, comprehensive simulation, and standardized assessment. Users can improve training efficiency by previewing before entering practice. Furthermore, this embodiment achieves dynamic matching between task content and learner's ability through iterative updates of the ability profile at each stage, truly achieving adaptive learning and enabling individualized instruction.

[0088] Furthermore, a four-stage closed-loop learning process is constructed, encompassing preview, practice, simulation, and formal testing. AI-driven dynamic content generation, fine-grained ability diagnosis, and personalized path recommendation mechanisms are introduced at each stage. By building a learner ability profile model, the system analyzes their performance data at each stage in real time, dynamically adjusting the training content and difficulty of subsequent stages to achieve closed-loop management of the entire process from knowledge input to ability output.

[0089] Figure 6This is a schematic block diagram of an intelligent training device based on adaptive learning provided in an embodiment of this application. Figure 6 As shown, corresponding to the above-described intelligent training method based on adaptive learning, this application also provides an intelligent training device 600 based on adaptive learning. This intelligent training device 600 includes units for executing the above-described intelligent training method based on adaptive learning, and can be configured in an intelligent training system based on adaptive learning. Specifically, please refer to... Figure 6 The intelligent training device 600 based on adaptive learning includes a transceiver unit 601 and a processing unit 602, wherein: Transceiver unit 601 is used to acquire pre-study instructions; Processing unit 602 is configured to: enter the pre-study phase in response to a user-triggered pre-study instruction; collect user pre-study behavior data on the pre-study materials via transceiver unit 601; update the user's basic ability profile based on the pre-study behavior data to obtain a first ability profile; enter the practice phase in response to a user-triggered practice instruction; generate personalized practice tasks based on the first ability profile, the personalized practice tasks including multiple single-round question-and-answer tasks; collect first operation data of the user on the personalized practice tasks via transceiver unit 601; update the first ability profile based on the first operation data to obtain a second ability profile; and respond to a user-triggered... The simulation phase begins upon receiving a simulated instruction. Based on the second competency profile and a preset customer profile, a simulated training task is generated. This simulated training task is a multi-turn dialogue task. The transceiver unit 601 collects the user's second operation data for the simulated training task and updates the second competency profile based on this data to obtain a third competency profile. In response to a user-triggered examination instruction, the examination phase begins. Based on the third competency profile and the current training objective, a standardized examination task is generated. The transceiver unit 601 collects the user's third operation data for the standardized examination task and generates a competency assessment report based on this third operation data.

[0090] In some embodiments, the transceiver unit 601 is further configured to acquire unstructured training materials corresponding to the current training task; The processing unit 602 is further configured to use natural language processing technology to extract key entities, entity relationships, and business rules from the unstructured training materials to construct a domain knowledge graph; to perform importance classification and cognitive dependency ranking of knowledge point nodes in the domain knowledge graph using a graph centrality algorithm to form a knowledge skeleton; to call a preset standardized teaching template and fill the knowledge point information in the knowledge skeleton into the corresponding slots of the template to obtain the first initial material; to perform pedagogical transcription of the first initial material using a large model to obtain the second initial material; and to perform multimodal structured assembly of the second initial material to generate the pre-study material, wherein the pre-study material includes at least one of knowledge point cards and flowcharts.

[0091] In some embodiments, the pre-study behavior data includes pre-study completion rate, saved content, question markers, and note content; when the processing unit 602 performs the step of updating the user's basic ability profile based on the pre-study behavior data to obtain a first ability profile, it is specifically used for: Based on the pre-study completion rate, the saved content, the question marks, and the notes, the user's weaknesses regarding the pre-study materials are determined. The basic capability profile is updated based on the weaknesses to obtain the first capability profile.

[0092] In some embodiments, the first operation data includes the answer content corresponding to each single-round question-and-answer task, the user's emotional state, and the user's body language; when the processing unit 602 performs the step of updating the first ability profile based on the first operation data to obtain the second ability profile, it is specifically used for: Based on the content of the answer, the user's emotional state, and the user's body language, a multi-dimensional answer score is determined, which assesses the completeness of knowledge points, the logic of the speech, the emotional expressiveness, and the expressiveness of the actions. The first ability profile is updated based on the multi-dimensional response score to obtain the second ability profile.

[0093] In some embodiments, after performing the step of determining a multi-dimensional answer score based on the answer content, the user's emotional state, and the user's body language regarding the completeness of knowledge points, the logicality of the speech, the expressiveness of emotions, and the expressiveness of actions, the processing unit 602 is further configured to: Based on the multidimensional response score, response suggestions are generated.

[0094] In some embodiments, when the processing unit 602 performs the step of generating a simulation training task based on the second capability profile and a preset customer profile, it is specifically used for: Based on the second capability profile, a target simulation scenario is selected from a preset scenario library; The simulation training task is generated based on the target simulation scenario, the preset user profile, and the preset random events.

[0095] In some embodiments, after performing the step of collecting the third operational data of the user for the standardized test task, the processing unit 602 is further configured to: The user is subjected to cheating detection based on the third operation data, and the cheating detection includes voiceprint recognition detection and screen monitoring detection.

[0096] In summary, this embodiment of the application constructs a four-stage progressive learning process, realizing a complete closed loop from knowledge preview to single-point practice, comprehensive simulation, and standardized assessment. Users can improve training efficiency by previewing before entering practice. Furthermore, this embodiment achieves dynamic matching between task content and learner's ability through iterative updates of the ability profile at each stage, truly achieving adaptive learning and enabling individualized instruction.

[0097] It should be noted that those skilled in the art can clearly understand that the specific implementation process of the above-mentioned intelligent training device based on adaptive learning and each unit can be referred to the corresponding description in the foregoing method embodiments. For the sake of convenience and brevity, it will not be repeated here.

[0098] The aforementioned intelligent training device based on adaptive learning can be implemented as a computer program, which can, for example... Figure 7 It runs on the computer device shown.

[0099] Please see Figure 7 , Figure 7 This is a schematic block diagram of a computer device provided in an embodiment of this application. The computer device 700 is a user terminal or server in intelligent training based on adaptive learning; See Figure 7 The computer device 700 includes a processor 702, a memory, and a network interface 705 connected via a system bus 701. The memory may include a non-volatile storage medium 703 and internal memory 704.

[0100] The non-volatile storage medium 703 may store an operating system 7031 and a computer program 7032. The computer program 7032 includes program instructions that, when executed, cause the processor 702 to perform an intelligent training method based on adaptive learning.

[0101] The processor 702 provides computing and control capabilities to support the operation of the entire computer device 700.

[0102] The internal memory 704 provides an environment for the execution of the computer program 7032 in the non-volatile storage medium 703. When the computer program 7032 is executed by the processor 702, the processor 702 can execute an intelligent training method based on adaptive learning.

[0103] This network interface 705 is used for network communication with other devices. Those skilled in the art will understand that... Figure 7 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device 700 to which the present application is applied. The specific computer device 700 may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0104] The processor 702 is used to run a computer program 7032 stored in the memory to perform the following steps: In response to a user-triggered pre-study instruction, the system enters the pre-study phase, collects the user's pre-study behavior data for the pre-study materials, and updates the user's basic ability profile based on the pre-study behavior data to obtain a first ability profile. In response to the user-triggered practice instruction, the user enters the practice phase and generates a personalized practice task based on the first ability profile. The personalized practice task includes multiple single-round question-and-answer tasks. Collect the user's first operation data for the personalized practice task, and update the first ability profile based on the first operation data to obtain the second ability profile; In response to the user-triggered simulation command, the simulation phase is entered, and a simulation training task is generated based on the second capability profile and the preset customer profile. The simulation training task is a multi-turn dialogue task. Collect the user's second operation data for the simulated training task, and update the second capability profile based on the second operation data to obtain a third capability profile; In response to the examination instruction triggered by the user, the examination phase is entered, and a standardized examination task is generated based on the third competency profile and the current training objectives. Collect the user's third operational data for the standardized test task, and generate the user's competency assessment report based on the third operational data.

[0105] It should be understood that in the embodiments of this application, the processor 702 may be a central processing unit (CPU), or it may be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor.

[0106] It will be understood by those skilled in the art that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program includes program instructions and can be stored in a storage medium, which is a computer-readable storage medium. The program instructions are executed by at least one processor in the computer system to implement the process steps of the embodiments of the above methods.

[0107] Therefore, this application also provides a storage medium. This storage medium can be a computer-readable storage medium. The storage medium stores a computer program, wherein the computer program includes program instructions. When executed by a processor, the program instructions cause the processor to perform the following steps: In response to a user-triggered pre-study instruction, the system enters the pre-study phase, collects the user's pre-study behavior data for the pre-study materials, and updates the user's basic ability profile based on the pre-study behavior data to obtain a first ability profile. In response to the user-triggered practice instruction, the user enters the practice phase and generates a personalized practice task based on the first ability profile. The personalized practice task includes multiple single-round question-and-answer tasks. Collect the user's first operation data for the personalized practice task, and update the first ability profile based on the first operation data to obtain the second ability profile; In response to the user-triggered simulation command, the simulation phase is entered, and a simulation training task is generated based on the second capability profile and the preset customer profile. The simulation training task is a multi-turn dialogue task. Collect the user's second operation data for the simulated training task, and update the second capability profile based on the second operation data to obtain a third capability profile; In response to the examination instruction triggered by the user, the examination phase is entered, and a standardized examination task is generated based on the third competency profile and the current training objectives. Collect the user's third operational data for the standardized test task, and generate the user's competency assessment report based on the third operational data.

[0108] The storage medium can be any computer-readable storage medium capable of storing program code, such as a USB flash drive, portable hard drive, read-only memory (ROM), magnetic disk, or optical disk.

[0109] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this application.

[0110] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative. For example, the division of each unit is merely a logical functional division, and there may be other division methods in actual implementation. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed.

[0111] The steps in the methods of this application embodiment can be adjusted, merged, or deleted according to actual needs. The units in the apparatus of this application embodiment can be merged, divided, or deleted according to actual needs. Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0112] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, a terminal, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application.

[0113] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in this application, and these modifications or substitutions should all be covered within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. An intelligent training method based on adaptive learning, characterized in that, include: In response to a user-triggered pre-study instruction, the system enters the pre-study phase, collects the user's pre-study behavior data for the pre-study materials, and updates the user's basic ability profile based on the pre-study behavior data to obtain a first ability profile. In response to the user-triggered practice instruction, the user enters the practice phase and generates a personalized practice task based on the first ability profile. The personalized practice task includes multiple single-round question-and-answer tasks. Collect the user's first operation data for the personalized practice task, and update the first ability profile based on the first operation data to obtain the second ability profile; In response to the user-triggered simulation command, the simulation phase is entered, and a simulation training task is generated based on the second capability profile and the preset customer profile. The simulation training task is a multi-turn dialogue task. Collect the user's second operation data for the simulated training task, and update the second capability profile based on the second operation data to obtain a third capability profile; In response to the examination instruction triggered by the user, the examination phase is entered, and a standardized examination task is generated based on the third competency profile and the current training objectives. Collect the user's third operational data for the standardized test task, and generate the user's competency assessment report based on the third operational data.

2. The method according to claim 1, characterized in that, Before collecting the user's pre-study behavior data regarding the pre-study materials, the method further includes: Obtain the unstructured training materials corresponding to the current coaching task; Natural language processing technology is used to extract key entities, entity relationships, and business rules from the unstructured training materials to construct a domain knowledge graph; The graph centrality algorithm is used to classify the importance of knowledge nodes in the domain knowledge graph and sort them by cognitive dependence to form a knowledge skeleton. By calling a preset standardized teaching template, the knowledge point information in the knowledge skeleton is filled into the corresponding slots of the template to obtain the first initial material; The first initial material is transcribed into a teaching format using a large model to obtain the second initial material; The second initial material is subjected to multimodal structured assembly to generate the pre-study material, which includes at least one of knowledge point cards and flowcharts.

3. The method according to claim 1, characterized in that, The pre-study behavior data includes pre-study completion rate, saved content, question marks, and note content; updating the user's basic ability profile based on the pre-study behavior data to obtain a first ability profile includes: Based on the pre-study completion rate, the saved content, the question marks, and the notes, the user's weaknesses regarding the pre-study materials are determined. The basic capability profile is updated based on the weaknesses to obtain the first capability profile.

4. The method according to claim 1, characterized in that, The first operational data includes the answers corresponding to each single-round question-and-answer task, the user's emotional state, and the user's body language. The step of updating the first capability profile based on the first operation data to obtain the second capability profile includes: Based on the content of the answer, the user's emotional state, and the user's body language, a multi-dimensional answer score is determined for the completeness of knowledge points, the logic of the speech, the emotional expressiveness, and the expressiveness of the actions. The first ability profile is updated based on the multi-dimensional response score to obtain the second ability profile.

5. The method according to claim 4, characterized in that, After determining the multi-dimensional response score based on the content of the response, the user's emotional state, and the user's body language, considering the completeness of knowledge points, the logic of the speech, the emotional expressiveness, and the expressiveness of the actions, the method further includes: Based on the multidimensional response score, response suggestions are generated.

6. The method according to claim 1, characterized in that, The simulation training task generated based on the second capability profile and the preset customer profile includes: Based on the second capability profile, a target simulation scenario is selected from a preset scenario library; The simulation training task is generated based on the target simulation scenario, the preset user profile, and the preset random events.

7. The method according to claim 1, characterized in that, After collecting the user's third operational data for the standardized test task, the method further includes: The user is subjected to cheating detection based on the third operation data, and the cheating detection includes voiceprint recognition detection and screen monitoring detection.

8. An intelligent training device based on adaptive learning, characterized in that, include: The transceiver unit is used to receive pre-study instructions; The processing unit is used to enter the pre-study stage in response to a pre-study instruction triggered by the user, collect the user's pre-study behavior data for the pre-study materials through the transceiver unit, and update the user's basic ability profile based on the pre-study behavior data to obtain a first ability profile. In response to the user-triggered practice instruction, the user enters the practice phase and generates a personalized practice task based on the first ability profile. The personalized practice task includes multiple single-round question-and-answer tasks. The transceiver unit collects the user's first operation data for the personalized practice task, and updates the first ability profile based on the first operation data to obtain a second ability profile. In response to the user-triggered simulation command, the simulation phase is entered, and a simulation training task is generated based on the second capability profile and the preset customer profile. The simulation training task is a multi-turn dialogue task. The transceiver unit collects the user's second operation data for the simulated training task, and updates the second competency profile based on the second operation data to obtain a third competency profile; in response to the user's triggered examination instruction, the examination phase is entered, and a standardized examination task is generated based on the third competency profile and the current training objective; The transceiver unit collects the user's third operational data for the standardized examination task and generates the user's competency assessment report based on the third operational data.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the intelligent training method based on adaptive learning as described in any one of claims 1-7.

10. A storage medium, characterized in that, The storage medium stores a computer program, which includes program instructions that, when executed by a processor, cause the processor to perform the intelligent training method based on adaptive learning as described in any one of claims 1-7.

Images