Historical self-learning system based on virtual intelligent tutoring

Abstract

The invention relates to the technical field of intelligent education, in particular to a history self-learning system based on virtual intelligent tutoring, which comprises a screening classification unit, a recommendation learning module, a learning unit, a deepening module, a feedback module and a secondary deepening module. Wherein the screening and classifying unit is used for classifying learners into a function and benefit oriented type and an emotional resonance type by adopting a learner classifying algorithm based on POS (Point of Sale) annotation and a random forest; the recommendation learning module is used for recommending different historical learning contents to the learner according to the classification result and a self-learning mechanism; the learning unit is used for providing various learning modes for the learner; the deepening module is used for deepening the learning of the learner in an intelligent tutoring mode; the feedback module is used for feeding back a learning condition to the learner and optimizing a self-learning mechanism according to a feedback result; and the secondary deepening module is used for reminding the learner to review and customizing review questions for the learner. Therefore, the problems that in the prior art, courses lack personalized adaptation, learning interaction is not deep, feedback lags and the like are solved.

Description

Technical Field

[0001] This application relates to the field of intelligent education technology, and particularly to a historical self-study system based on virtual intelligent tutoring. Background Art

[0002] Traditional historical education mainly relies on teachers' lectures in the classroom, with textbooks as the core carrier, and teaches historical knowledge to students through blackboard writing, oral narration, etc. With the progress of technology, multimedia teaching has gradually become popular, and teachers have begun to use materials such as pictures and videos to enrich teaching content. The rise of online education platforms has also provided students with more channels for independent study of history, and students can obtain learning resources such as historical course videos and e-books on the platforms. However, the existing historical education systems and methods have many limitations, especially in the integration with current advanced deep learning, artificial intelligence and other cutting-edge technologies.

[0003] In related technologies, historical education systems mainly rely on the combination of traditional education models and digital technologies. In terms of the presentation of content, it is mainly through the combination of text and multimedia, and historical events and character introductions are presented through e-textbooks, PPTs or web pages, supplemented by pictures, videos, etc. to enhance intuitiveness. In the design of courses, chapters are mainly divided according to the time axis or topics, and learners browse the content in sequence, lacking a personalized jump mechanism; in terms of interaction methods, it is mainly one-way knowledge transfer, and learners passively receive, and the interaction is limited to leaving messages in the discussion area; in the after-class test, memory-based knowledge is assessed through some fixed multiple-choice questions, fill-in-the-blank questions and other question types, lacking dynamic adjustment ability; in the evaluation of learning effect feedback, it is mostly based on stage tests, and surface data such as learning duration and question answering accuracy rate are statistically analyzed, lacking the excavation of learning behaviors and the depth of knowledge mastery, and unable to capture learning weaknesses in real time. Summary of the Invention

[0004] This application provides a historical self-study system based on virtual intelligent tutoring to solve the problems in the prior art such as the lack of personalized adaptation of courses, the lack of in-depth learning interaction, and the lag of feedback.

[0005] The embodiment of the first aspect of this application provides a historical self-learning system based on virtual intelligent tutoring, including: a screening and classification unit, a recommended learning module, a learning unit, a deepening module, a feedback module, and a secondary deepening module; among them, the screening and classification unit is used to classify learners, adopt a learner classification algorithm based on POS tagging and random forest, input the information of learners and test result data into the classification model to output classification results of utilitarian orientation type and emotional resonance type; the recommended learning module is used to recommend historical learning content in different directions and difficulties for learners according to the classification of learners and a self-learning mechanism based on the integration of reinforcement learning and meta-learning; the learning unit is used to provide various learning methods for learners; the deepening module is used to deepen the learning of learners through intelligent tutoring, and the intelligent tutoring includes virtual historical scene deduction, virtual character debate-style question-and-answer tutoring, 3D cultural relic display, and historical association; the feedback module is used to feedback the learning situation to learners and optimize the self-learning mechanism according to the feedback results; the secondary deepening module is used to remind learners to review and customize review questions for learners.

[0006] Optionally, the screening and classification unit includes: an acquisition unit, a test unit, and an analysis and classification module. Among them, the acquisition unit is used to acquire the information of learners, and the information of learners includes age, gender, and current educational background information; the test unit is used to design multiple-choice questions related to historical learning to test learners. Each option in the multiple-choice questions contains text descriptions. Suppose the number of multiple-choice questions is M. For the i-th multiple-choice question (1 i M), the number of its options is , record the answer option data O={ , , , ,}, where represents the option number of the learner for the i-th multiple-choice question, and at the same time collect the text content T={ } of each option, represents the text content of the j-th option (1 ≤ j ≤ Ni) of the i-th multiple-choice question; the analysis and classification module is used to judge the type of learners according to the information of learners and test results.

[0007] Optionally, the analysis and classification module includes a POS tagging unit, a feature extraction unit, and a random forest classification unit. Among them, the POS tagging unit is used to preprocess each option of the multiple-choice question, use a word segmentation tool to segment the option text into single words to obtain a part-of-speech set ={ }, where is the option text The number of words after word segmentation, perform part-of-speech tagging on the words after word segmentation, and for each word (1 ) assign corresponding part-of-speech tags ; The feature extraction unit is used to extract multi-dimensional part-of-speech features, and construct a comprehensive feature vector F by combining the information of the learner and the coding features of the answer options themselves. Among them, the multi-dimensional part-of-speech features include part-of-speech frequency features, part-of-speech combination features, and keyword part-of-speech features. Among them, The part-of-speech frequency feature is used for the option text , set the part-of-speech set as P = { , , , } (1 ) frequency The calculation formula is , Among them, is the number of words with part-of-speech in the option text , is the number of words after word segmentation of the option text ; The part-of-speech combination feature is used for the option text , the frequency of the part-of-speech combination c is the proportion of the number of times this part-of-speech combination appears in the option to the total number of q-gram combinations; The keyword part-of-speech feature is used to set the keyword set as K = { , , , }, for the keyword in the option text (1 ≤ s ≤ S), record its part-of-speech and the part-of-speech of the words within its surrounding range ; The random forest classification unit is used to input the comprehensive feature vector F into the random forest classifier, optimize the performance of the classifier by adjusting parameters such as the number of decision trees and the maximum depth, and output the learner classification result. The classification result is utilitarian-oriented and emotional resonance type.

[0008] Optionally, the recommendation learning module includes: a meta-learning unit, a reinforcement learning unit, and a content recommendation unit, wherein the meta-learning unit is used to collect historical user learning behavior data, analyze the learning strategies of different types of learners, and generate initial recommendation strategy parameters for different types of users; the reinforcement learning unit is used to dynamically optimize the recommendation strategy using a deep Q reinforcement learning network based on the learner portrait and the current recommended content characteristics and learning feedback; the content recommendation unit is used to execute content recommendation, build a candidate content library according to the classification type, calculate the content recommendation score according to the fusion strategy, and match the recommended content.

[0009] Optionally, the learning unit includes: a learning method unit, which is used to provide learners with multiple learning methods, including course following, text study, and voice guidance.

[0010] Optionally, the deepening module includes: a virtual historical scene interpretation unit, a virtual character debate question and answer coaching unit, a 3D cultural relic display unit, and a historical association unit, wherein the virtual historical scene interpretation unit is used to construct an immersive virtual historical scene based on historical records and digital technology to give learners an immersive situational experience; the virtual character debate question and answer coaching unit is used to create a virtual historical character role library to guide learners to think deeply about historical viewpoints and controversial issues in the form of debate and question and answer; the 3D cultural relic display unit is used to display historical relics with three-dimensional visualization and age slider technology, and can display the state changes of cultural relics according to the sliding of the age, so as to assist learners in understanding the culture and historical value behind the cultural relics; the historical association unit is used to explore multi-dimensional associations to help learners build systematic cognition, and the multi-dimensional associations include climate associations, domestic and foreign contemporary event associations, and interdisciplinary associations; Optionally, the feedback module includes: a progress analysis unit, an error analysis unit, a knowledge mastery analysis unit, a learning efficiency evaluation analysis unit, an attentiveness analysis unit, and a self-learning mechanism optimization unit. The progress analysis unit is used to track the completion of the learner's historical learning plan, compare the preset goals with the actual progress, and provide a visual progress report. The error analysis unit is used to identify the errors in the learner's answering questions and operations, analyze the types and roots of the errors, and provide targeted improvement suggestions. The knowledge mastery analysis unit is used to quantitatively evaluate the depth and breadth of the learner's understanding of historical knowledge points and construct a dynamic knowledge mastery map. The learning efficiency evaluation analysis unit is used to analyze the amount of knowledge acquired by the learner per unit time and the input-output ratio, and identify inefficient learning modes. The attentiveness analysis unit is used to monitor the learner's attention state through multimodal data, identify distracted periods, and adjust learning strategies. The multimodal data includes the page stay duration, the intervals of keyboard and mouse operations, etc. The self-learning mechanism optimization unit is used to summarize the analysis results of the progress analysis unit, the error analysis unit, the knowledge mastery analysis unit, the learning efficiency evaluation analysis unit, and the attentiveness analysis unit, and drive the iteration of the learning strategies of the recommended learning module.

[0011] Optionally, the secondary deepening module includes: a reminder review unit and a customized question unit. The reminder review unit is used to dynamically trigger review reminders according to the forgetting curve to improve the knowledge retention rate. The customized question unit is used to generate differentiated review questions based on the learner's knowledge weaknesses and learning goals to achieve precise leak detection and make up for deficiencies.

[0012] The second aspect of the embodiments of the present application provides a historical self-learning method based on virtual intelligent tutoring, including: obtaining the information of the learner and the test data information. The information of the learner includes age, gender, and current educational level information. The test data information includes the learner's answering option data and the text content of each option. Construct a learner classification model based on POS tagging and random forest according to the information of the learner and the test data information, and classify the learner into a utility-oriented type and an emotional resonance type according to the classification model. Build a self-learning mechanism that integrates reinforcement learning and meta-learning, and recommend different historical learning contents for the learner according to the classification of the learner and the self-learning mechanism. Provide a learning method for learning according to the recommended different historical learning contents, and provide an intelligent tutoring method for deepening learning to obtain the learning situation of the learner. The learning method includes course following, text study, and voice guided reading. The intelligent tutoring includes virtual historical scene deduction, virtual character debate-style question-and-answer tutoring, 3D cultural relic display, and historical association. Analyze and feedback to the learner according to the learning situation, and dynamically remind the learner to review based on the forgetting curve, and customize review questions for the learner.

[0013] A third aspect embodiment of the present application provides an electronic device, including: a memory, a processor, and a computer program stored on the memory and executable on the processor, where the processor executes the program to implement the historical self-learning method based on virtual intelligent tutoring as in the above embodiments.

[0014] The beneficial effects obtained by the present invention are as follows: The embodiment of the present application uses a learner classification algorithm based on POS tagging and random forest to mine the part-of-speech semantic features of the learner's answer option text, and combines basic information to achieve motivation-oriented classification, which is convenient for recommending more preferred learning content for learners; based on the fusion mechanism of meta-learning and reinforcement learning, it can provide common recommendation strategies according to the learner type, and can also be optimized in real time according to individual feedback, realizing self-learning dynamic recommendation of learning content and improving the matching efficiency of learning resources; through intelligent tutoring methods such as virtual historical scene deduction, virtual character debate-style question and answer tutoring, and 3D cultural relic display, historical learning is upgraded from the memory level to situational exploration. Through multi-dimensional historical associations, it helps learners construct a knowledge association map in many aspects, improves the knowledge retention rate, and realizes in-depth interaction; the feedback module integrates multi-dimensional data such as progress, errors, and concentration, drives the dynamic iteration of the self-learning mechanism, and realizes the closed-loop optimization of the self-learning recommendation mechanism; the secondary deepening module combines the forgetting curve and customized questions to improve the consolidation efficiency of learners' knowledge points. Thus, the problems of lack of personalized adaptation of courses, lack of in-depth learning interaction, and lagging feedback in the prior art are solved.

[0015] Additional aspects and advantages of the present application will be given in part in the following description, will become apparent in part from the following description, or will be understood through the practice of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The above and / or additional aspects and advantages of the present application will become apparent and easy to understand from the following description of the embodiments in conjunction with the drawings, where: Figure 1 is a schematic structural diagram of a historical self-learning system based on virtual intelligent tutoring according to an embodiment of the present application; Figure 2 is a schematic diagram of a screening and classification unit according to an embodiment of the present application; Figure 3 is a schematic diagram of a recommended learning module according to an embodiment of the present application; Figure 4 is a schematic diagram of a deepening module according to an embodiment of the present application; Figure 5 is a schematic diagram of a feedback module according to an embodiment of the present application; Figure 6 is a schematic diagram of a secondary deepening module according to an embodiment of the present application; Figure 7 Flow chart of a historical self - learning method based on virtual intelligent tutoring according to an embodiment of the present application; Figure 8 Schematic structural diagram of an electronic device according to an embodiment of the present application. Detailed implementation manners

[0017] The embodiments of the present application will be described in detail below. The examples of the embodiments are shown in the accompanying drawings, where the same or similar reference numerals denote the same or similar elements or elements with the same or similar functions from beginning to end. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to explain the present application, and should not be construed as a limitation to the present application.

[0018] The historical self - learning system based on virtual intelligent tutoring according to the embodiments of the present application will be described below with reference to the accompanying drawings. Aiming at the problems in the prior art mentioned in the above - mentioned background technology, such as lack of personalized adaptation of courses, lack of in - depth learning interaction, and lagging feedback, the present application provides a historical self - learning system based on virtual intelligent tutoring. In this system, based on the learner classification algorithm of POS tagging and random forest, the part - of - speech semantic features of the learner's answer option text are mined, and combined with basic information, motivation - oriented classification is realized, which is convenient for recommending more preferred learning content for learners; based on the fusion mechanism of meta - learning and reinforcement learning, it can provide common recommendation strategies according to the learner type, and can also be optimized in real - time according to individual feedback, realizing self - learning dynamic recommendation of learning content and improving the matching efficiency of learning resources; through intelligent tutoring methods such as virtual historical scene deduction, virtual character debate - style question - answering tutoring, and 3D cultural relic display, historical learning is upgraded from the memory level to situational exploration. Through multi - dimensional historical associations, it helps learners construct a knowledge association map in many aspects, improves the knowledge retention rate, and realizes in - depth interaction; the feedback module integrates multi - dimensional data such as progress, errors, and concentration, driving the dynamic iteration of the self - learning mechanism, and realizing the closed - loop optimization of the self - learning recommendation mechanism; the secondary deepening module combines the forgetting curve and customized questions to improve the consolidation efficiency of learners' knowledge points. Thus, the problems in the prior art, such as lack of personalized adaptation of courses, lack of in - depth learning interaction, and lagging feedback, are solved.

[0019] Specifically, Figure 1 Schematic diagram of the composition of the historical self - learning system based on virtual intelligent tutoring provided by the embodiments of the present application.

[0020] As Figure 1 shown, the historical self - learning system 10 based on virtual intelligent tutoring includes: A screening and classification unit 100, a recommended learning module 200, a learning unit 300, a deepening module 400, a feedback module 500, and a secondary deepening module 600.

[0021] Among them, the screening and classification unit 100 is used to classify learners. By adopting a learner classification algorithm based on POS tagging and random forest, the information of learners and the test result data are input into the classification model to output classification results of utility-oriented and emotional resonance types; the recommended learning module 200 is used to recommend historical learning content in different directions and difficulties for learners according to the classification of learners and a self-learning mechanism based on the integration of reinforcement learning and meta-learning; the learning unit 300 is used to provide learners with various learning methods; the deepening module 400 is used to deepen the learning of learners through intelligent tutoring, and the intelligent tutoring includes virtual historical scene deduction, virtual character debate-style Q&A tutoring, 3D cultural relic display, and historical association; the feedback module 500 is used to feedback the learning situation to learners and optimize the self-learning mechanism according to the feedback results; the secondary deepening module 600 is used to remind learners to review and customize review questions for learners.

[0022] It can be understood that through the screening and classification unit in the embodiment of the present application, different types of learners can be accurately identified. Based on this, the recommended learning module provides content that meets the learning styles and goals of learners. Whether it is utility-oriented or emotional resonance type learners, they can obtain suitable learning resources to meet the needs of individual differences. By providing various learning methods, learners can flexibly choose according to their own habits and scenarios, improving the participation and convenience of learning. The deepening module helps learners deeply understand historical knowledge from different angles through various intelligent tutoring methods, constructs a systematic knowledge system, and cultivates learners' critical thinking and historical analysis abilities. Through learning result feedback and secondary deepening, it helps learners discover problems, solve problems in a timely manner, optimize learning methods, and improve knowledge retention rate.

[0023] In the embodiment of the present application, the screening and classification unit 100 includes: as Figure 2 shown, an acquisition unit, a test unit, and an analysis and classification module.

[0024] Among them, the acquisition unit is used to acquire the information of learners, and the information of learners includes age, gender, and current educational background information; the test unit is used to design multiple-choice questions related to historical learning to test learners. Each option in the multiple-choice questions contains text descriptions. Let the number of multiple-choice questions be M. For the i-th multiple-choice question (1 i M), the number of its options is , record the answer option data O = { , , , }, where represents the option number of the learner for the i-th multiple-choice question, and at the same time collect the text content T = { } of each option. Represents the text content of the j-th option (1 ≤ j ≤ Ni) of the i-th multiple-choice question; the analysis and classification module is used to determine the type of the learner based on the learner's information and test results.

[0025] Specifically, the learner is 18 years old, male, and currently has a high school education. When testing the learner, 15 questions are selected for the test, and each question has 4 options. Taking the first question as an example, the question description is: "When studying the 'Silk Road', what will you prioritize?" Option 1 is to master the main cities connected by the Silk Road, commodity circulation, and the change of the ancient economic pattern; Option 2 is to explore the touching stories behind cross-cultural exchanges such as the eastward spread of Buddhism and Xuanzang's westward journey; Option 3 is to memorize the iconic events such as Zhang Qian's mission to the Western Regions and the Dunhuang Grottoes and their historical status; Option 4 is to appreciate the art and vicissitudes of the Silk Road civilization through the murals of the Mogao Grottoes and the Loulan ruins. If the learner selects 3, then record as 3, ={ }={Master the main cities connected by the Silk Road, commodity circulation, and the change of the ancient economic pattern, explore the touching stories behind cross-cultural exchanges such as the eastward spread of Buddhism and Xuanzang's westward journey, memorize the iconic events such as Zhang Qian's mission to the Western Regions and the Dunhuang Grottoes and their historical status, appreciate the art and vicissitudes of the Silk Road civilization through the murals of the Mogao Grottoes and the Loulan ruins}. Each multiple-choice question is recorded in sequence to form the final answer option data O and option text T, and the data is passed into the analysis and classification module to obtain the classification result as utilitarian-oriented.

[0026] It can be understood that in this embodiment of the application, by collecting the basic information of the learner such as age, gender, and education level and the behavioral data such as answer options and text semantics, it is convenient to analyze the learning motivation of the learner from multiple dimensions. Structuring the answer data is convenient for the training of the subsequent random forest model.

[0027] In this embodiment of the application, the analysis and classification module includes a POS tagging unit, a feature extraction unit, and a random forest classification unit. Among them, The POS tagging unit is used to preprocess each multiple-choice question option, and use a word segmentation tool to segment the option text into individual words to obtain the part-of-speech set ={ }, where is the number of words after word segmentation of the option text. Perform part-of-speech tagging on the words after word segmentation, and for each word (1 ) assign the corresponding part-of-speech tag ; The feature extraction unit is used to extract multi-dimensional part-of-speech features, and construct a comprehensive feature vector F by combining the information of the learner and the encoding features of the answer options themselves. Among them, the multi-dimensional part-of-speech features include part-of-speech frequency features, part-of-speech combination features, and keyword part-of-speech features. Among them, The part-of-speech frequency feature is used for the option text , let the part-of-speech set be P = { , , , }, the frequency of part-of-speech (1 ) is calculated by the formula as , where is the number of words with part-of-speech in the option text , is the number of words after word segmentation of the option text ; The part-of-speech combination feature is used for the option text , the frequency of part-of-speech combination c is the proportion of the number of occurrences of this part-of-speech combination in the option to the total number of q-gram combinations; The keyword part-of-speech feature is used to set the keyword set as K = { , , , }, for the keyword in the option text (1 ≤ s ≤ S), record its part-of-speech and the part-of-speech of the words within its surrounding range ; The random forest classification unit is used to input the comprehensive feature vector F into the random forest classifier, optimize the classifier performance by adjusting parameters such as the number of decision trees and the maximum depth, and output the learner classification result, and the classification result is utilitarian-oriented and emotional resonance type.

[0028] Specifically, taking the above-mentioned 18-year-old male high school diploma learner and the third option of the "Silk Road" multiple-choice question, "Memorize landmark events such as Zhang Qian's mission to the Western Regions and the Dunhuang Grottoes and their historical status", as an example, it is tokenized and part-of-speech tagged through the POS tagging unit to obtain the part-of-speech set W = ["Memorize", "Zhang Qian", "mission to the Western Regions", "Dunhuang", "Grottoes", "etc.", "landmark", "event", "and", "its", "historical", "status"], and the part-of-speech tags P = ["VB" (verb), "NNP" (proper noun), "VB", "NNP", "NN" (noun), "CC" (conjunction), "JJ (adjective)", "NN", "CC", "NN", "NN"]. Through the feature extraction unit, part-of-speech frequency features, part-of-speech combination features, and keyword part-of-speech features are obtained. For example, the part-of-speech frequency of NN is 6 / 11 (including NNP), the part-of-speech frequency of NNP is 2 / 11, the part-of-speech frequency of VB is 2 / 11, and the part-of-speech combinations are VB_NNP, NNP_VB, VB_NNP, NNP_NN, NN_CC, CC_JJ, JJ_NN, NN_CC, CC_NN, NN_NN. Then the part-of-speech frequency of VB_NNP is 2 / 10, that is, 0.2. Assuming that one of the keywords is "landmark event" (JJ_NN), its surrounding part-of-speech is CC. According to the learner information and the extracted multi-dimensional features, a comprehensive feature vector F is constructed and input into the random forest classification unit, and the output classification result is utilitarian-oriented.

[0029] It can be understood that through the POS tagging method and multi-dimensional feature extraction of the option text in the embodiments of the present application, the semantics of the test options are deeply mined, the potential learning motivation of the learners is captured, and the subjective answering preferences are transformed into objective feature vectors, and combined with the random forest model for training, realizing the accurate classification of historical learners at the motivation level and laying a foundation for subsequent personalized recommendations.

[0030] In the embodiments of the present application, the recommended learning module 200 includes: as Figure 3 shown, a meta-learning unit, a reinforcement learning unit, and a content recommendation unit.

[0031] Among them, the meta-learning unit is used to collect the learning behavior data of historical users, analyze the learning strategies of different types of learners, and generate initial recommendation strategy parameters for different types of users; the reinforcement learning unit is used to dynamically optimize the recommendation strategy according to the learner profile, the current recommended content features, and the learning feedback using a deep Q reinforcement learning network; the content recommendation unit is used to execute content recommendations, construct a candidate content library according to the classification type, calculate the content recommendation score according to the fusion strategy, and match the recommended content.

[0032] Specifically, the formula for the meta-learning unit to generate the initial recommendation strategy parameters is:

[0033]

[0034] where 1 ≤ i ≤ N, 1 ≤ j ≤ M, N is the number of historical users, and M is the number of user types. is the number of users of a certain type. is the learning behavior data of a certain historical user, and k is the feature dimension of the learning behavior data. is the mean vector of the behavior data of a certain type of user. is the initial recommendation strategy parameter of a certain type. It can be understood that the embodiments of the present application adopt meta-learning technology to generate an initial strategy using group experience, avoiding the cold start problem of recommendations for new users due to insufficient data. A deep Q reinforcement learning network is used to dynamically adjust the recommendation strategy according to real-time feedback such as dwell time and collection behavior. For example, if a learner frequently skips a certain type of content, the system will automatically reduce the weight of this type to improve the recommendation accuracy. Building a content library by type and performing quantitative matching can effectively reduce invalid recommendations and optimize resource utilization.

[0035] In the embodiments of the present application, the learning unit 300 includes: a learning method unit.

[0036] The learning method unit is used to provide multiple learning methods for learners, and the learning methods include course following, text study, and voice guidance.

[0037] It can be understood that the embodiments of the present application provide multiple learning methods for learners, adapting to the learning preferences of different learners. Course following presents teaching content in video form, adapting to the visual learning needs. Text study provides annotation analysis to meet the preference for in-depth text analysis. Voice guidance is suitable for auditory learning and fragmented scenarios. The three learning methods can cooperate to strengthen the learning effect. First, establish a macroscopic understanding through course following, then analyze details through text study, and finally review and consolidate with voice guidance to improve the knowledge retention rate.

[0038] In the embodiments of the present application, the deepening module 400 includes: as Figure 4 shown, a virtual historical scene deduction unit, a virtual character debate-style Q&A tutoring unit, a 3D cultural relic display unit, and a historical association unit.

[0039] Among them, the virtual historical scenario deduction unit is used to construct an immersive virtual historical scenario based on historical records and digital technology, giving learners an immersive situational experience; the virtual character debate-style Q&A tutoring unit is used to create a virtual historical character library, guiding learners to deeply think about historical viewpoints and controversial issues in the form of debate and Q&A; the 3D cultural relic display unit is used to display historical cultural relics with three-dimensional visualization and chronological slider technology, and can display the state changes of cultural relics according to the sliding of the chronological order, assisting learners to understand the cultural and historical values behind the cultural relics; the historical association unit is used to excavate multi-dimensional associations to help learners build a systematic understanding, and the multi-dimensional associations include climate associations, domestic and foreign contemporary event associations, and interdisciplinary associations. Specifically, when learning about the "Industrial Revolution", the system constructs a virtual textile factory based on the historical archives and architectural drawings of Manchester, England in the 18th century. Learners can "enter" the workshop where the steam engines are roaring and observe the process of workers operating the spinning jenny.

[0040] Specifically, when discussing the topic "Was Shang Yang's Reform too severe?", the system creates virtual characters Shang Yang, Gan Long, and Sima Qian. Among them, Shang Yang is the supporter, Gan Long is the opponent, and Sima Qian is the neutral character. After the learner asks a question, Shang Yang argues from the perspective of "harsh laws are needed in troubled times", Gan Long refutes with "the people's resentment will arise", and Sima Qian supplements the historical evaluation in combination with "Records of the Grand Historian", guiding the learner to analyze the advantages and disadvantages of the reform.

[0041] Specifically, when a learner wants to learn about the cultural relic Simuwu Ding, 3D technology is used to model and display the Simuwu Ding. At the same time, through the chronological slider, it can be slid from the initial casting period (late Shang Dynasty) to the modern era. It can be seen the rust changes on the body of the ding caused by long-term burial and the damage marks at the time of excavation. Clicking on the pattern can display the cultural interpretation of "the taotie pattern symbolizes the royal power".

[0042] Specifically, when learning about "Zheng He's voyages to the Western Seas", the system conducts multi-dimensional associations. At the climate level, it shows the impact of the monsoon pattern in the Indian Ocean during the same period on the route selection; at the level of domestic and foreign events, it compares the timeline and navigation technology differences of the European Age of Discovery; at the interdisciplinary level, the system associates astronomy to understand the star navigation of the fleet and associates geography to understand the principles of monsoons and ocean currents. It helps learners build a systematic understanding from multiple perspectives.

[0043] It can be understood that the immersive virtual historical scenario constructed in the embodiments of the present application makes history concrete through multi-sensory stimulation, breaks the limitation of standard answers through role debate-style Q&A, cultivates the speculative ability and critical thinking of learners, displays cultural relics through three-dimensional visualization and chronological slider technology, makes static cultural relics dynamic, enhances learners' interest and sense of identity in cultural relics, and helps learners jump out of the isolated event cognition and understand the multi-dimensional driving factors of historical development by constructing multi-dimensional historical associations. Through various intelligent tutoring methods, fragmented history is transformed into an experiential, speculative, and associative learning ecosystem, solving the pain points of "isolated events and superficial understanding" in traditional history education.

[0044] In the embodiments of the present application, the feedback module 500 includes: as Figure 5 shown, a progress analysis unit, an error analysis unit, a knowledge mastery degree analysis unit, a learning efficiency evaluation analysis unit, a concentration analysis unit, and a self-learning mechanism optimization unit.

[0045] Among them, the progress analysis unit is used to track the completion of the learner's historical learning plan, compare the preset goal with the actual progress, and provide a visual progress report; the error analysis unit is used to identify the errors in the learner's answering questions and operations, analyze the error types and roots, and provide targeted improvement suggestions; the knowledge mastery degree analysis unit is used to quantitatively evaluate the depth and breadth of the learner's understanding of historical knowledge points and construct a dynamic knowledge mastery map; the learning efficiency evaluation analysis unit is used to analyze the amount of knowledge acquired by the learner per unit time and the input-output ratio, and identify inefficient learning modes; the concentration analysis unit is used to monitor the learner's attention state through multi-modal data, identify distraction periods and adjust learning strategies, and the multi-modal data includes page stay duration, keyboard and mouse operation intervals, etc.; the self-learning mechanism optimization unit is used to summarize the analysis results of the progress analysis unit, the error analysis unit, the knowledge mastery degree analysis unit, the learning efficiency evaluation analysis unit, the concentration analysis unit, and the self-learning mechanism optimization unit, and drive the iteration of the learning strategy of the recommended learning module.

[0046] It can be understood that the embodiments of the present application comprehensively monitor and evaluate the learning situation of learners by integrating five dimensions of progress analysis, error analysis, knowledge mastery degree, learning efficiency evaluation, and concentration analysis, provide targeted improvement suggestions and adjustment strategies, which helps learners discover and solve problems in a timely manner, optimize learning methods, and improve learning efficiency and quality. At the same time, inputting the evaluation results into the self-learning model can promote the optimization of the self-learning model, drive the iteration of the learning strategy of the recommended learning module, and realize the recommended learning closed-loop of "recommendation - learning - evaluation - optimization".

[0047] In the embodiments of the present application, the secondary deepening module 600 includes: as Figure 6 shown, a reminder review unit and a customized question unit.

[0048] Among them, the review reminder unit is used to dynamically trigger review reminders according to the forgetting curve to improve the knowledge retention rate; the customized question unit is used to generate differentiated review questions based on the learner's knowledge weaknesses and learning objectives to achieve precise leakage detection and make up for deficiencies.

[0049] It can be understood that the embodiments of the present application dynamically calculate the best review nodes based on the forgetting curve, remind learners to review in time, and help improve the memory retention rate of learners through the design of scientific review time nodes. The learning weaknesses are located through the information provided by the feedback unit, and personalized review questions are customized to precisely improve the learning short board.

[0050] According to the historical self-learning system based on virtual intelligent tutoring proposed by the embodiments of the present application, a learner classification algorithm based on POS tagging and random forest is used to mine the part-of-speech semantic features of the learner's answer option text, and combined with basic information, motivation-oriented classification is realized, which is convenient for recommending more preferred learning content for learners; based on the fusion mechanism of meta-learning and reinforcement learning, it can provide common recommendation strategies according to the learner type, and can also be optimized in real time according to individual feedback, realizing the self-learning dynamic recommendation of learning content and improving the learning resource matching efficiency; through intelligent tutoring methods such as virtual historical scene deduction, virtual character debate-style Q&A tutoring, and 3D cultural relic display, historical learning is upgraded from the memory level to situational exploration. Through multi-dimensional historical associations, it helps learners build a knowledge association map in many aspects, improves the knowledge retention rate, and realizes deep interaction; the feedback module integrates multi-dimensional data such as progress, errors, and concentration, drives the dynamic iteration of the self-learning mechanism, and realizes the closed-loop optimization of the self-learning recommendation mechanism; the secondary deepening module combines the forgetting curve with customized questions to improve the consolidation efficiency of learners' knowledge points. Thus, the problems of lack of personalized adaptation of courses, lack of in-depth learning interaction, and lagging feedback in the prior art are solved.

[0051] Secondly, a historical self-learning method based on virtual intelligent tutoring proposed according to the embodiments of the present application is described with reference to the accompanying drawings.

[0052] As Figure 7 shown, the historical self-learning method based on virtual intelligent tutoring includes the following steps: In step S101, the information of the learner and the test data information are obtained.

[0053] Among them, the information of the learner includes age, gender, current educational background information, and the test data information includes the learner's answer option data and the text content of each option.

[0054] Specifically, the learner is 26 years old, female, and has a bachelor's degree. When testing the learner, 15 questions are selected for the test, and each question has 4 options. Record the answer options of the 15 questions of the learner and the option text content of the 15 questions.

[0055] It can be understood that in the embodiment of the present application, by obtaining the physiological characteristics, educational background, and behavior data of the user, a learner portrait is constructed, providing materials for semantic mining and data support for subsequent classification and personalized recommendation.

[0056] In step S102, a learner classification model based on POS tagging and random forest is constructed according to the learner's information and test data information, and the learners are classified into utilitarian-oriented and emotional resonance types according to the classification model.

[0057] Specifically, each obtained answer option text is segmented, the segmented words are POS tagged, multi-dimensional part-of-speech features are extracted and combined with the learner's information to construct a comprehensive feature vector. The multi-dimensional part-of-speech features include part-of-speech frequency features, part-of-speech combination features, and keyword part-of-speech features. The comprehensive feature vector is input into a random forest classifier to obtain a classification result.

[0058] For example, a 26-year-old female learner with a bachelor's degree, when faced with the test question "If you were a young person during the War of Resistance Against Japanese Aggression, what would you prioritize?" selected the option text "Organize the masses to publicize the anti-Japanese ideology". After segmenting and POS tagging this option text, a part-of-speech set ["organize", "masses", "publicize", "anti-Japanese", "ideology"] is obtained, and the part-of-speech tags are ["VB" (verb), "NN" (noun), "VB", "JJ" (adjective), "NN"]. The part-of-speech frequency features, part-of-speech combination features, and keyword part-of-speech features are extracted in turn. For example, the part-of-speech frequency of VB is 2 / 5, that is, 0.4, the part-of-speech frequency of NN is 2 / 5, that is, 0.4, the part-of-speech frequency of JJ is 0.2, and the part-of-speech combination features are ["VB_NN", "NN_VB", "VB_JJ", "JJ_NN"], where the part-of-speech frequency of each is 0.25. Assuming that one of the keywords is "anti-Japanese ideology" (JJ_NN), its surrounding part-of-speech is VB. A comprehensive feature vector F is constructed according to the learner's information and the extracted multi-dimensional features and input into a random forest classifier, and the output classification result is emotional resonance type.

[0059] It can be understood that by performing POS part-of-speech tagging and multi-dimensional part-of-speech feature extraction on the answer option text in the embodiments of the present application, semantic and emotional information in the text can be captured, the internal learning motivation of learners can be deeply explored, rather than relying solely on surface information or simple behavior patterns, making the classification results better reflect the true needs and preferences of learners. The multi-dimensional part-of-speech features and comprehensive feature vectors can capture the complex relationships between words in the text. For example, the part-of-speech combination features can reflect the collocations and semantic associations between words. The random forest algorithm can process these high-dimensional and complex feature data, effectively discover the potential connections between features, and thus classify learners more accurately.

[0060] In step S103, a self-learning mechanism based on the integration of reinforcement learning and meta-learning is built, and different historical learning contents are recommended for learners according to their classification and the self-learning mechanism.

[0061] Specifically, a new user is determined to be utilitarian-oriented. According to meta-learning technology, the utilitarian group strategy is extracted from historical user data to generate initial recommendation parameters. For example, the content type is test point analysis, and the difficulty coefficient is low level, etc. The recommendation strategy is dynamically adjusted in real time according to the learner's learning behavior feedback by the reinforcement learning network. For example, if a learner skips the document "The Reign of the Jacobins" during learning but repeatedly watches the video "The Rise of Napoleon", then the weight of abstract theory content is reduced, and the priority of content combining character stories and test points is increased.

[0062] It can be understood that in the embodiments of the present application, meta-learning is used to generate an initial strategy using group experience, and new users can obtain basic adapted content without trial and error. By reinforcement learning, individual differences are captured in real time, and the recommendation strategy is dynamically adjusted, solving the pain points of "difficult cold start" and "weak adaptability" in traditional recommendations, improving the accuracy rate of recommendations, and helping to improve the learning efficiency of learners.

[0063] In step S104, according to the recommended different historical learning contents, learning methods are provided for learning, and an intelligent tutoring method is provided for in-depth learning to obtain the learning situation of learners.

[0064] Among them, the learning methods include course following, text study, and voice guided reading. The intelligent tutoring methods include virtual historical scene deduction, virtual character debate-style Q&A tutoring, 3D cultural relic display, and historical association.

[0065] Specifically, when recommending the learning content of the British Industrial Revolution to a utilitarian-oriented learner, learning methods such as course following, text study, and voice-guided reading are provided. The learner, who wants to quickly master the core knowledge points, chooses the course-following method to study. Subsequently, the learner uses the virtual character debate-style Q&A tutoring function to ask the question "Did the Industrial Revolution do more harm than good?" Debates are generated with the roles of capitalists, worker representatives, and environmental enthusiasts respectively, and the impacts of the Industrial Revolution are understood from multiple perspectives. Then, the historical correlation function is used to study the impact of the climate at that time on the Industrial Revolution and to understand technologies such as steam engines in combination with physics, constructing a cognition of the Industrial Revolution from multiple angles.

[0066] It can be understood that the embodiments of the present application recommend different historical learning contents and provide diverse learning methods for learning, which can meet the habits and scenario needs of different learners. With the intelligent tutoring means, it can help learners observe and think about problems from different angles, cultivate historical thinking ability and imagination, effectively activate learning interest at the same time, and significantly deepen the learning effect.

[0067] In step S105, the learner is analyzed and fed back according to the learning situation, and the learner is dynamically reminded to review based on the forgetting curve, and review questions are customized for the learner.

[0068] It can be understood that through the accurate analysis and feedback of the learner's learning situation in the embodiments of the present application, the weak points of knowledge and learning preferences can be intuitively presented. The dynamic review reminder based on the forgetting curve can intervene scientifically at the key memory nodes to avoid knowledge forgetting. Combined with the customized review questions, it not only makes up for the short board but also deepens the understanding. This trinity mechanism transforms passive learning into precise reinforcement, significantly improving the memory retention rate and knowledge application ability, and at the same time adapting to the individual learning rhythm, making historical review efficient and targeted.

[0069] According to the historical self - learning method based on virtual intelligent tutoring proposed in the embodiments of the present application, a learner classification algorithm based on POS tagging and random forest is used to mine the part - of - speech semantic features of the learner's answer option text. Combining with basic information, it realizes motivation - oriented classification, which is convenient for recommending more preferred learning content for learners; based on the fusion mechanism of meta - learning and reinforcement learning, it can provide common recommendation strategies according to the learner type and can also be optimized in real - time according to individual feedback, realizing the self - learning dynamic recommendation of learning content and improving the learning resource matching efficiency; through intelligent tutoring methods such as virtual historical scene deduction, virtual character debate - style question - answering tutoring, and 3D cultural relic display, historical learning is upgraded from the memory level to situational exploration. Through multi - dimensional historical associations, it helps learners construct a knowledge association graph in many aspects, improves the knowledge retention rate, and realizes deep interaction; the feedback module integrates multi - dimensional data such as progress, errors, and concentration, driving the dynamic iteration of the self - learning mechanism and realizing the closed - loop optimization of the self - learning recommendation mechanism; the secondary deepening module combines the forgetting curve and customized questions to improve the consolidation efficiency of learners' knowledge of key points. Thus, the problems of lack of personalized adaptation of courses, lack of in - depth learning interaction, and lagging feedback in the existing technology are solved.

[0070] The following will specifically elaborate on the historical self - learning method based on virtual intelligent tutoring, and the content is as follows: S1: Obtain the information of the learner and the information of the test data.

[0071] For example, it is obtained that learner A is 16 years old, female, and the current education level is high school. When testing the learner, 15 questions are selected for the test, and each question has 4 options. Record the answer options of the learner's 15 questions and the option text content of the 15 questions.

[0072] S2: Construct a learner classification model based on POS tagging and random forest according to the information of the learner and the test data information, and classify the learners into utilitarian - oriented type and emotional resonance type according to the classification model.

[0073] For example, when learner A faces the test question "What is the measure in the Meiji Restoration that best reflects learning from the West?", the option text "Advocate civilization and enlightenment, and learn Western culture" is selected. After segmenting the option text, POS part - of - speech tagging and part - of - speech labels are performed, and the part - of - speech frequency features, part - of - speech combination features, and keyword part - of - speech features are extracted in turn. According to the learner information and the extracted multi - dimensional features, a comprehensive feature vector F is constructed and input into the random forest classifier, and the classification result is the emotional resonance type.

[0074] S3: Build a self - learning mechanism based on the fusion of reinforcement learning and meta - learning, and recommend different historical learning content for learners according to the classification of learners and the self - learning mechanism.

[0075] For example, learner A is a new user and is determined to be an emotional resonance type. Group experience is obtained from emotional resonance type users according to meta-learning technology, and initial recommended content is pushed, such as learning content like "Artistic Changes in the Meiji Period", "Meiji Restoration: Women in the Era of Change", "The Arrival of the Black Ships: The Moment Japan's National Gates Were Opened", etc. According to the reinforcement learning network, the recommendation strategy is dynamically adjusted in real-time based on the learner's learning behavior. For example, learner A carefully read the graphic materials of "Meiji Restoration: Women in the Era of Change" and also wrote down a lot of her own insights beside it. However, for the documentary "The Arrival of the Black Ships: The Moment Japan's National Gates Were Opened", although she watched it once, she didn't show any particular interest. Therefore, the recommendation weight of documentaries on relatively macroscopic historical events was reduced, and the recommendation priority of content focusing on the individual fates of characters and social and cultural details was increased.

[0076] S4: Provide learning methods for learning according to different recommended historical learning content, and provide intelligent tutoring methods for in-depth learning to obtain the learning situation of the learner.

[0077] For example, learner A chose the course following method to learn the recommended content, and then used the virtual historical scene deduction function to enter the virtual scene of Tokyo School of Fine Arts to participate in the study of Western oil painting techniques. She also used the 3D cultural relic display function to view the Western-style clothing designs in the Meiji period.

[0078] S5: Analyze and feedback to the learner based on the learning situation, and dynamically remind the learner to review based on the forgetting curve, and customize review questions for the learner.

[0079] For example, conduct progress analysis, error analysis, knowledge mastery level, learning efficiency evaluation, and concentration analysis on the learning situation of learner A, and provide targeted improvement suggestions and adjustment strategies. Based on the forgetting curve, push review reminders one day, three days, seven days, and fifteen days after learning, and customize corresponding review questions.

[0080] In summary, through the learner classification model based on POS tagging and random forest in the embodiments of the present application, the semantic information of the learner's test options is deeply mined, and the learners are classified at the motivation level, so as to recommend learning content to the learners according to their learning motivation. Meta-learning is used to accelerate the generation of new user strategies and reduce the cold start time. The reinforcement learning network dynamically optimizes the transfer strategy according to the user's real-time feedback to form personalized recommendations. Deepening learning through intelligent tutoring methods can significantly enhance the immersion, speculative nature, and knowledge integration efficiency of historical learning. Analyzing and feedback on the learning situation helps learners discover problems, solve problems in a timely manner, and optimize learning methods. The review reminder based on the forgetting curve can help learners improve the knowledge retention rate, learning efficiency, and quality.

[0081] Figure 8Schematic diagram of the structure of the electronic device provided by the embodiment of the present application. The electronic device may include: A memory 801, a processor 802, and a computer program stored on the memory 801 and executable on the processor 802.

[0082] When the processor 802 executes the program, it implements the historical self-learning method of virtual intelligent tutoring provided in the above embodiment.

[0083] Furthermore, the electronic device further includes: A communication interface 803 for communication between the memory 801 and the processor 802.

[0084] The memory 801 is used to store a computer program executable on the processor 802.

[0085] The memory 801 may include a high-speed RAM memory, and may also include a non-volatile memory, such as at least one disk memory.

[0086] If the memory 801, the processor 802, and the communication interface 803 are implemented independently, the communication interface 803, the memory 801, and the processor 802 can be interconnected through a bus and communicate with each other. The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc. For the sake of representation, Figure 8 only a thick line is used to represent it in the figure, but it does not mean that there is only one bus or one type of bus.

[0087] Optionally, in a specific implementation, if the memory 801, the processor 802, and the communication interface 803 are integrated on a chip, the memory 801, the processor 802, and the communication interface 803 can communicate with each other through an internal interface.

[0088] The processor 802 may be a Central Processing Unit (CPU), or an Application Specific Integrated Circuit (ASIC), or one or more integrated circuits configured to implement the embodiment of the present application.

[0089] In the description of this specification, the descriptions referring to terms such as "one embodiment", "some embodiments", "example", "specific example", or "some examples" mean that the specific features, structures, materials, or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of this application. In this specification, the schematic descriptions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described can be combined in any one or N embodiments or examples in a suitable manner. In addition, without contradiction, those skilled in the art can combine and combine the different embodiments or examples described in this specification and the features of different embodiments or examples.

[0090] In addition, the terms "first" and "second" are used for descriptive purposes only and cannot be construed as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features. Thus, the features defined with "first" and "second" may explicitly or implicitly include at least one of such features. In the description of this application, the meaning of "N" is at least two, such as two, three, etc., unless otherwise specifically defined.

[0091] Any process or method description shown in a flowchart or described in other ways herein can be understood to represent a module, segment, or part of code including one or more N executable instructions for implementing a customized logical function or process. And the scope of the preferred embodiments of this application includes additional implementations, where the functions can be executed in a substantially simultaneous manner or in a reverse order according to the involved functions, rather than in the order shown or discussed, which should be understood by those skilled in the technical field to which the embodiments of this application belong.

[0092] It should be understood that each part of this application can be implemented by hardware, software, firmware, or a combination thereof. In the above embodiments, the N steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, any one or a combination of the following techniques well known in the art can be used: discrete logic circuits having logic gate circuits for implementing logical functions on data signals, application specific integrated circuits having appropriate combinational logic gate circuits, programmable gate arrays (PGAs), field programmable gate arrays (FPGAs), etc.

[0093] Those of ordinary skill in the technical field of this application can understand that all or part of the steps carried by the method for implementing the above embodiments can be completed by instructing relevant hardware through a program, and the program can be stored in a computer-readable storage medium. When the program is executed, it includes one or a combination of the steps of the method embodiments.

Claims

1. A historical self-learning system based on virtual intelligent tutoring, characterized in that, It includes the following steps: Screening classification units, recommended learning modules, learning units, deepening modules, feedback modules, and secondary deepening modules; among them, The screening classification unit is used to classify learners. By using a learner classification algorithm based on POS tagging and random forest, the information of learners and test result data are input into the classification model to output classification results of utilitarian - oriented and emotional resonance - oriented types; The recommended learning module is used to recommend historical learning content in different directions and difficulties for learners according to the classification of learners and a self - learning mechanism based on the integration of reinforcement learning and meta - learning; The learning unit is used to provide learners with various learning methods; The deepening module is used to deepen learners' learning through intelligent tutoring. The intelligent tutoring includes virtual historical scene deduction, virtual character debate - style question - answering tutoring, 3D cultural relic display, and historical association; The feedback module is used to feedback the learning situation to learners and optimize the self - learning mechanism according to the feedback results; The secondary deepening module is used to remind learners to review and customize review questions for learners.

2. The historical self-learning system based on virtual intelligent tutoring according to claim 1, characterized in that The screening classification unit includes: an acquisition unit, a test unit, and an analysis and classification module. Among them, The acquisition unit is used to acquire the information of learners. The information of learners includes age, gender, and current educational level information; The test unit is used to test learners by designing multiple-choice questions related to historical learning. Each option in the multiple-choice questions contains a text description. Let the number of multiple-choice questions be M. For the i-th multiple-choice question (1 i ≤ i ≤ M), the number of its options is Ni. Record the answer option data of the learner O = { oi1 , oi2 , ...}, where oi represents the option number of the learner for the i-th multiple-choice question. At the same time, collect the text content of each option T = { ti1 }, where tij represents the text content of the j-th option (1 ≤ j ≤ Ni) of the i-th multiple-choice question; The analysis and classification module is used to judge the type of learners according to the information of learners and test results.

3. The historical self-learning system based on virtual intelligent tutoring according to claim 2, characterized in that, The analysis and classification module includes a POS tagging unit, a feature extraction unit, and a random forest classification unit. Among them, The POS tagging unit is used to preprocess each multiple-choice question option, and uses a word segmentation tool to segment the option text into individual words, obtaining a set of part-of-speech tags ={ }, where is the option text is the number of words after word segmentation. Perform part-of-speech tagging on the words after word segmentation, and assign corresponding part-of-speech tags to each word (1 ); The feature extraction unit is used to extract multi - dimensional part - of - speech features, and construct a comprehensive feature vector F by combining the information of learners and the encoding features of the answer options themselves. Among them, the multi - dimensional part - of - speech features include part - of - speech frequency features, part - of - speech combination features, and keyword part - of - speech features. Among them, The part-of-speech frequency feature is used for the option text , let the part-of-speech set be P = { , , , }, the frequency of part-of-speech (1 ) is The calculation formula is , Among them, is the text of the option in which the part of speech is the number of words is the text of the option the number of words after word segmentation; The part-of-speech combination feature is used for option text , the frequency of part-of-speech combination c is the proportion of the number of occurrences of the part-of-speech combination in the option to the total number of q-gram combinations; The keyword part-of-speech features are used to set the keyword set as K = { , , , }. For the keywords in the option text (1 ≤ s ≤ S), record its part of speech and the part of speech of the words within its surrounding range ; The random forest classification unit is used to input the comprehensive feature vector F into the random forest classifier, optimize the performance of the classifier by adjusting parameters such as the number of decision trees and the maximum depth, and output the learner classification results. The classification results are utilitarian - oriented and emotional resonance - oriented types.

4. The historical self-learning system based on virtual intelligent tutoring according to claim 1, characterized in that The recommended learning module includes: a meta - learning unit, a reinforcement learning unit, and a content recommendation unit. Among them, the meta - learning unit is used to collect the learning behavior data of historical users, analyze the learning strategies of different types of learners, and generate initial recommendation strategy parameters for different types of users; the reinforcement learning unit is used to dynamically optimize the recommendation strategy by using a deep Q - reinforcement learning network according to the learner portrait, the characteristics of the current recommended content, and the learning feedback; the content recommendation unit is used to execute content recommendation, construct a candidate content library according to the classification type, calculate the content recommendation score according to the fusion strategy, and match the recommended content.

5. The historical self-learning system based on virtual intelligent tutoring according to claim 1, characterized in that, The learning unit includes: a learning method unit. The learning method unit is used to provide learners with various learning methods. The learning methods include course following, text study, and voice guided reading.

6. The historical self-learning system based on virtual intelligent tutoring according to claim 1, wherein The deepening module includes: a virtual historical scenario deduction unit, a virtual character debate-style Q&A tutoring unit, a 3D cultural relic display unit, and a historical correlation unit. Among them, the virtual historical scenario deduction unit is used to construct an immersive virtual historical scenario based on historical records and digital technology, giving learners an immersive situational experience; the virtual character debate-style Q&A tutoring unit is used to create a virtual historical character library, guiding learners to deeply think about historical viewpoints and controversial issues in the form of debate and Q&A; the 3D cultural relic display unit is used to display historical cultural relics with three-dimensional visualization and chronological slider technology, and can display the state changes of cultural relics according to the sliding of the chronology, assisting learners to understand the cultural and historical values behind the cultural relics; the historical correlation unit is used to excavate multi-dimensional correlations, helping learners build a systematic cognition, and the multi-dimensional correlations include climate correlation, domestic and foreign contemporary event correlation, and interdisciplinary correlation.

7. The historical self-learning system based on virtual intelligent tutoring according to claim 1, wherein, The feedback module includes: a progress analysis unit, an error analysis unit, a knowledge mastery degree analysis unit, a learning efficiency evaluation and analysis unit, a concentration analysis unit, and a self-learning mechanism optimization unit. Among them, the progress analysis unit is used to track the completion of the learner's historical learning plan, compare the preset goal with the actual progress, and provide a visual progress report; the error analysis unit is used to identify the errors in the learner's answering questions and operations, analyze the error types and roots, and provide targeted improvement suggestions; the knowledge mastery degree analysis unit is used to quantitatively evaluate the depth and breadth of the learner's understanding of historical knowledge points and construct a dynamic knowledge mastery map; the learning efficiency evaluation and analysis unit is used to analyze the amount of knowledge acquired by the learner per unit time and the input-output ratio, and identify inefficient learning modes; the concentration analysis unit is used to monitor the learner's attention state through multi-modal data, identify distraction periods and adjust learning strategies, and the multi-modal data includes page stay duration, keyboard and mouse operation intervals, etc.; the self-learning mechanism optimization unit is used to summarize the analysis results of the progress analysis unit, the error analysis unit, the knowledge mastery degree analysis unit, the learning efficiency evaluation and analysis unit, and the concentration analysis unit, and drive the iteration of the learning strategies of the recommended learning module.

8. The historical self-learning system based on virtual intelligent tutoring according to claim 1, characterized in that, The secondary deepening module includes: a reminder review unit and a customized question unit. Among them, the reminder review unit is used to dynamically trigger review reminders according to the forgetting curve to improve the knowledge retention rate; the customized question unit is used to generate differentiated review questions based on the learner's knowledge weaknesses and learning goals to achieve precise leakage detection and supplementation.

9. A historical self-learning method based on virtual intelligent tutoring, characterized in that, including: Obtain the learner's information and test data information, where the learner's information includes age, gender, and current educational background information, and the test data information includes the learner's answering option data and the text content of each option. Construct a learner classification model based on POS tagging and random forest according to the learner's information and test data information, and classify learners into utilitarian-oriented and emotional resonance types according to the classification model. Build a self-learning mechanism based on the integration of reinforcement learning and meta-learning, and recommend different historical learning contents for learners according to the classification of learners and the self-learning mechanism. Recommend different historical learning contents according to the above, provide learning methods for learning, and provide intelligent tutoring methods for in-depth learning to obtain the learning situation of learners. Among them, the learning methods include course following, text study, and voice guided reading, and the intelligent tutoring includes virtual historical scene deduction, virtual character debate-style Q&A tutoring, 3D cultural relic display, and historical association; Analyze and feedback to the learners according to the learning situation and dynamically remind the learners to review based on the forgetting curve, and customize review questions for the learners.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, The program is executed by a processor to implement the historical self-learning method based on virtual intelligent tutoring as described in claim 9.

Images