An intelligent learning companion method, device, storage medium and equipment

Abstract

The present application discloses an intelligent learning companion method, device, storage medium and equipment. The method includes: first, obtaining the first learning information and the second learning information of the target user, and then using the first learning information of the target user, in combination with the first prompt, inputting it into the LLM to obtain the user profile of the target user output by the model; then using the user profile of the target user, the second learning information of the target user and the agent information, in combination with the second prompt, inputting it into the LLM to obtain the learning strategy for the target user output by the model; and then according to the learning strategy, through the LLM, the agents are collaboratively called to realize the intelligent learning companion for the target user. Thus, the timeliness and accuracy of the entire learning companion process are effectively improved, and further the learning companion effect for the target user is effectively improved, and the learning experience of the target user is also enhanced.

Description

Technical Field

[0001] This application relates to the field of natural language processing technology, and particularly to an intelligent companion learning method, device, storage medium and equipment. Background Art

[0002] With the rapid development of educational technology, people's demands for personalized learning and teaching students in accordance with their aptitude are becoming stronger and stronger. However, many existing auxiliary learning methods do not really meet the users' learning demands in a timely manner and achieve the goal of reducing the burden and improving the efficiency of learning.

[0003] Currently, the existing methods for assisting users in learning generally include three types: the first is to use self-searching resources / tools learning products for auxiliary learning; the second is to use personalized auxiliary learning tool products based on learning situation modeling for auxiliary learning; the third is a synchronous planning learning type of auxiliary learning method under expert knowledge. However, none of these existing methods for assisting users in learning have truly introduced the concept of companion learning. They only reflect the accompaniment of a cold training machine, rather than playing the role of a learning partner who can customize a learning plan for users and answer questions in a timely manner during the process. Therefore, the ideal companion learning effect cannot be achieved, and the users' learning experience will also be reduced. Summary of the Invention

[0004] The main purpose of the embodiments of this application is to provide an intelligent companion learning method, device, storage medium and equipment, which can first construct a personalized portrait of the user based on a large language model (LLM), and then use a process self-reflection mechanism to effectively improve the companion learning effect for the user, thereby enhancing the user's learning experience.

[0005] The embodiments of this application provide an intelligent companion learning method, including:

[0006] Obtain the first learning information and the second learning information of the target user;

[0007] Use the first learning information of the target user, combined with the first prompt, and input it into the large language model to obtain the user portrait of the target user output by the model;

[0008] Use the user portrait of the target user, the second learning information of the target user, and the agent information, combined with the second prompt, and input it into the large language model to obtain the learning strategy for the target user output by the model;

[0009] According to the learning strategy, the large language model is used to collaboratively call the agent to achieve companion learning for the target user.

[0010] In a possible implementation, the first learning information of the target user includes the current learning status of the target user, the historical learning information of the target user, and the scenario information where the target user is located; using the first learning information of the target user, combined with the first prompt instruction "prompt", input it into the large language model to obtain the user portrait of the target user output by the model, including:

[0011] Structurally process the historical learning information of the target user, and use the obtained processing result combined with the first prompt instruction "prompt", input it into the large language model to obtain the historical portrait of the target user output by the model;

[0012] Use the current learning status of the target user, the historical portrait of the target user, and the scenario information where the target user is located, combined with the updated first prompt instruction "prompt", input it into the large language model to obtain the instant portrait of the target user output by the model;

[0013] Use the historical portrait of the target user and the instant portrait of the target user to construct the overall user portrait of the target user.

[0014] In a possible implementation, the second learning information of the target user includes the learning needs proposed by the target user; the learning needs proposed by the target user are determined according to at least one round of conversation content between the large language model and the target user.

[0015] In a possible implementation, using the user portrait of the target user, the second learning information of the target user, and the agent information, combined with the second prompt instruction "prompt", input it into the large language model to obtain the learning strategy for the target user output by the model, including:

[0016] Use the user portrait of the target user, the learning needs proposed by the target user, the scenario information where the target user is located, and the agent information, combined with the second prompt instruction "prompt", input it into the large language model to obtain the learning strategy for the target user output by the model.

[0017] In a possible implementation, according to the learning strategy, through the large language model to make a collaborative call to the agent to achieve accompanying learning for the target user, including:

[0018] According to the learning strategy, through the large language model to call each agent to achieve the main process of accompanying learning for the target user; and in the process of implementing the refined process of accompanying learning, through the mutual call between each agent, achieve the collaborative call of each agent to complete the entire process of accompanying learning for the target user.

[0019] In a possible implementation, the method further includes:

[0020] During the accompanying learning process, according to the feedback information of the target user, using the process self-reflection mechanism, update the user profile of the target user;

[0021] Using the updated user profile of the target user, update the learning strategy of the target user, and using the updated learning strategy, once again cooperate with the large language model to call the intelligent agent to achieve intelligent accompanying learning for the target user until the learning needs proposed by the target user are met.

[0022] In a possible implementation, during the accompanying learning process, according to the feedback information of the target user, using the process self-reflection mechanism, updating the user profile of the target user includes:

[0023] During the accompanying learning process, determine the target learning resources to be recommended according to the learning strategy; and judge whether the relevance between the target learning resources and the learning needs and user profile of the target user meets the preset requirements;

[0024] If so, recommend the target learning resources to the target user, and conduct real-time interaction with the target user according to the learning participation degree, interaction during the participation process, various learning behaviors and learning attitudes of the target user, so as to recommend the adjusted target learning resources to the target user according to the feedback information of the target user, and update the relevant content in the user profile of the target user;

[0025] If not, confirm the adjusted target learning resources through a secondary interaction with the target user; and recommend the adjusted target learning resources to the target user, and update the relevant content in the user profile of the target user;

[0026] Analyze the learning results of the target user for the target learning resources and / or the adjusted target learning resources, and update the user profile of the target user according to the analysis results.

[0027] In a possible implementation, the analysis results include the insufficient completion degree of the learning content of the target user for the target learning resources and / or the target learning resources and the error cause analysis results of the wrong questions.

[0028] In a possible implementation, the current learning state of the target user includes at least one of the target user's current expression, the text proposed by the target user, and / or the verbal and behavioral expressions uttered; the historical learning information of the target user includes at least one of the target user's historical behavior sequence, interaction sequence, answer content, process performance, and results; the scenario information where the target user is located includes at least one of a review scenario, a preview scenario, a preparation scenario for an exam, and a holiday learning scenario.

[0029] The embodiments of the present application also provide an intelligent companion learning device, including:

[0030] An acquisition unit, configured to acquire the first learning information and the second learning information of the target user;

[0031] A first input unit, configured to input the first learning information of the target user, combined with a first prompt instruction prompt, into a large language model to obtain the user portrait of the target user output by the model;

[0032] A second input unit, configured to input the user portrait of the target user, the second learning information of the target user, and the agent information, combined with a second prompt instruction prompt, into the large language model to obtain the learning strategy for the target user output by the model;

[0033] A companion learning unit, configured to, according to the learning strategy, collaboratively call the agent through the large language model to implement companion learning for the target user.

[0034] In a possible implementation, the first learning information of the target user includes the current learning state of the target user, the historical learning information of the target user, and the scenario information where the target user is located; the first input unit includes:

[0035] A first input subunit, configured to structurally process the historical learning information of the target user, and input the obtained processing result, combined with a first prompt instruction prompt, into the large language model to obtain the historical portrait of the target user output by the model;

[0036] A second input subunit, configured to input the current learning state of the target user, the historical portrait of the target user, and the scenario information where the target user is located, combined with an updated first prompt instruction prompt, into the large language model to obtain the instant portrait of the target user output by the model;

[0037] A construction subunit, configured to construct the overall user portrait of the target user by using the historical portrait and the instant portrait of the target user.

[0038] In a possible implementation, the second learning information of the target user includes the learning requirements proposed by the target user; the learning requirements proposed by the target user are determined based on at least one round of conversation content between the large language model and the target user.

[0039] In a possible implementation, the second input unit is specifically configured to:

[0040] Input the user profile of the target user, the learning requirements proposed by the target user, the scenario information where the target user is located, and the agent information, combined with the second prompt instruction prompt, into the large language model to obtain the learning strategy for the target user output by the model.

[0041] In a possible implementation, the accompanying learning unit is specifically configured to:

[0042] According to the learning strategy, call each agent through the large language model to implement the main process of accompanying learning for the target user; and in the process of implementing the refined process of accompanying learning, through the mutual calls between each agent, realize the collaborative call of each agent, and complete the entire process of accompanying learning for the target user.

[0043] In a possible implementation, the device further includes:

[0044] An update unit, configured to update the user profile of the target user by using a process self-reflection mechanism according to the feedback information of the target user during the accompanying learning process;

[0045] A call unit, configured to update the learning strategy of the target user by using the updated user profile of the target user, and use the updated learning strategy to call the agents collaboratively again through the large language model to implement intelligent accompanying learning for the target user until the learning requirements proposed by the target user are met.

[0046] In a possible implementation, the update unit includes:

[0047] A judgment subunit, configured to determine the target learning resources to be recommended according to the learning strategy during the accompanying learning process; and judge whether the relevance between the target learning resources and the learning requirements and user profile of the target user meets the preset requirements;

[0048] The first update subunit is configured to, if it is determined that the relevance between the target learning resource and the learning needs and user profile of the target user meets the preset requirements, recommend the target learning resource to the target user, and perform real-time interaction with the target user according to the learning participation degree of the target user, the interaction during the participation process, and various learning behaviors and learning attitudes, so as to recommend the adjusted target learning resource to the target user according to the feedback information of the target user, and update the relevant content in the user profile of the target user;

[0049] The second update subunit is configured to, if it is determined that the relevance between the target learning resource and the learning needs and user profile of the target user does not meet the preset requirements, confirm the adjusted target learning resource through secondary interaction with the target user; and recommend the adjusted target learning resource to the target user, and update the relevant content in the user profile of the target user;

[0050] The third update subunit is configured to analyze the learning results of the target user for the target learning resource and / or the adjusted target learning resource, and update the user profile of the target user according to the analysis results.

[0051] In a possible implementation manner, the analysis results include that the completion degree of the learning content of the target user for the target learning resource and / or the target learning resource is insufficient and the error cause analysis results of the wrong questions.

[0052] In a possible implementation manner, the current learning state of the target user includes at least one of the current expression of the target user, the text proposed by the target user, and / or the verbal and behavioral expressions issued; the historical learning information of the target user includes at least one of the historical behavior sequence, interaction sequence, answer content, process performance, and result of the target user; the scenario information where the target user is located includes at least one of a review scenario, a preview scenario, a preparation scenario for an exam, and a holiday learning scenario.

[0053] An embodiment of the present application further provides an intelligent learning companion device, including: a processor, a memory, and a system bus;

[0054] The processor and the memory are connected through the system bus;

[0055] The memory is used to store one or more programs, and the one or more programs include instructions, and when the instructions are executed by the processor, the processor is caused to execute any one of the implementation manners of the above intelligent learning companion method.

[0056] The embodiments of the present application also provide a computer-readable storage medium, in which instructions are stored. When the instructions run on a terminal device, the terminal device is caused to execute any one of the implementation manners of the above-mentioned intelligent learning assistance method.

[0057] The embodiments of the present application also provide a computer program product. When the computer program product runs on a terminal device, the terminal device is caused to execute any one of the implementation manners of the above-mentioned intelligent learning assistance method.

[0058] An intelligent learning assistance method, device, storage medium and device provided by the embodiments of the present application first obtain the first learning information and the second learning information of a target user; then use the first learning information of the target user, in combination with the first prompt instruction prompt, and input it into the LLM to obtain the user portrait of the target user output by the model; then use the user portrait of the target user, the second learning information of the target user and the agent information, in combination with the second prompt instruction prompt, and input it into the LLM to obtain the learning strategy for the target user output by the model; then according to the learning strategy, the agent is collaboratively called through the LLM to realize the learning assistance for the target user.

[0059] It can be seen that since the present application first accurately constructs the full-scenario portrait of the target user based on the LLM using the first learning information of the target user (such as the historical learning information, current learning status and scene information of the target user, etc.), and then determines the learning strategy based on the full-scenario portrait of the target user, the second learning information of the target user (such as the learning requirements proposed by the target user, etc.) and the agent information, and according to the learning strategy, through the collaborative call of the agent and the process self-reflection mechanism, the intelligent learning assistance for the target user is realized, effectively improving the timeliness and accuracy of the entire learning assistance process, thereby improving the learning assistance effect for the target user, and further enhancing the learning experience of the target user. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0061] Figure 1 It is a schematic flowchart of an intelligent learning assistance method provided by an embodiment of the present application;

[0062] Figure 2 It is a schematic diagram of the overall implementation process of the intelligent learning assistance provided by an embodiment of the present application;

[0063] Figure 3 A process example diagram for updating the user profile of a target user according to the feedback information of the target user and using the process self-reflection mechanism provided by an embodiment of the present application;

[0064] Figure 4 A schematic diagram of the composition of an intelligent learning companion device provided by an embodiment of the present application. Detailed implementation manners

[0065] With the rapid development of educational technology, to meet people's demands for personalized learning and individualized instruction, more and more learning aids have emerged in people's lives. However, many current products that assist users in learning still tend to fixedly allocate learning resources to users under a given learning syllabus. Of course, with the evolution of technology, personalized recommendation of learning resources has been realized through modeling historical learning situations. But basically, current products that assist users in learning are all passive for users to accept, there is no free interaction for users and no timely response or insight into the real needs of users, and the emotions in the learning process of users cannot be experienced, resulting in users feeling that the learning process is very boring.

[0066] Specifically, products that assist users in learning basically fall into the following three categories:

[0067] Category 1: Retrieval tool / resource type. Such learning products mainly provide a large number of resources to facilitate users' own screening and training. The main feature of the products is rich resources, and users need to have a strong awareness of independent screening. The corresponding technical solutions are mainly based on tag search.

[0068] Category 2: Auxiliary learning tool type. Such methods mainly build a user learning situation profile based on the historical learning situation sequence of users, and then predict the user's future knowledge point mastery situation through probability, so as to recommend subsequent learning paths. The focus is on prediction technology and students need to use it frequently to enrich the historical learning situation.

[0069] Category 3: Synchronous planning learning type. The main idea is to distribute relevant resources to users regularly according to the teaching syllabus or relevant teaching progress, and the process may be adaptively adjusted in combination with user levels, etc. But essentially, it is still a combination of resource allocation and tool scheduling, and the overall perception given to users is to do tasks according to a plan.

[0070] It can be seen that the current existing products related to assisting users in learning are basically concentrated in self-retrieving resources / tools, personalized learning based on learning situation modeling, and planning learning under expert knowledge, etc. The concept of accompanying learning has not really emerged. The accompanying learning here more reflects learning companionship, enhancing the user's learning confidence and sense of achievement during the learning process, making the user feel that the learning product (such as a learning machine) is not a cold training machine but a learning partner that understands me, tailors a learning plan for me, and answers questions in a timely manner during the process. Therefore, the existing methods and products for assisting users in learning cannot achieve the ideal accompanying learning effect and will also reduce the user's learning experience.

[0071] Moreover, the applicant has also found that in recent years, large language models (LLMs) have provided powerful reasoning and outstanding performance in dealing with complex human emotions and unstructured interactions, breaking through the bottleneck of the inability of traditional deep learning / machine learning to reach human-like capabilities. At the same time, along with the rapid development of agent technologies under various LLMs, not only can the teaching strategy be dynamically adjusted according to the learning situation and feedback of students, but also more user-friendly learning support can be provided by simulating the behavior of human teachers. Agents can also better integrate and analyze various data in the student's learning process, such as learning habits, preferences, and performance, so as to provide deeper teaching insights for teachers (such as LLMs can be understood as). This data-driven approach can help teachers (such as LLMs) formulate teaching plans more effectively and conduct personalized teaching according to the needs of students.

[0072] Therefore, this application proposes to combine advanced LLM technology to achieve a more user-friendly and intelligent accompanying learning process for users to solve the above defects. Specifically, this application provides an intelligent accompanying learning method. First, the first learning information and the second learning information of the target user are obtained; then, using the first learning information of the target user, combined with the first prompt instruction prompt, it is input into the LLM to obtain the user profile of the target user output by the model; then, using the user profile of the target user, the second learning information of the target user, and the agent information, combined with the second prompt instruction prompt, it is input into the LLM to obtain the learning strategy for the target user output by the model; then, according to the learning strategy, the LLM is used to coordinately call the agent to achieve the accompanying learning of the target user.

[0073] It can be seen that in this application, based on the large language model (LLM), the first learning information of the target user (such as the target user's historical learning information, current learning state, and scene information, etc.) is used to accurately construct the full-scenario portrait of the target user. Then, based on the full-scenario portrait of the target user, the second learning information of the target user (such as the learning needs proposed by the target user, etc.) and the agent information, the learning strategy is determined. According to this learning strategy, through the collaborative call of the agent and the process self-reflection mechanism, the intelligent companion learning for the target user is realized, effectively improving the timeliness and accuracy of the entire companion learning process, thereby improving the companion learning effect for the target user, and further enhancing the learning experience of the target user.

[0074] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of this application. Obviously, the described embodiments are some, but not all, of the embodiments of this application. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of this application without creative efforts shall fall within the scope of protection of this application.

[0075] First Embodiment

[0076] See Figure 1 , which is a schematic flowchart of an intelligent companion learning method provided in this embodiment. The method includes the following steps:

[0077] S101: Obtain the first learning information and the second learning information of the target user.

[0078] In this embodiment, any user (such as a student, a doctor, etc.) who uses the method provided in this embodiment to achieve intelligent companion learning is defined as the target user. Moreover, in order to realize a more user-friendly and intelligent companion learning process for the target user and improve the learning experience of the target user, this application first interacts with the target user in one or more rounds in real time based on the large language model (LLM), so as to obtain the learning needs proposed by the target user (that is, break down the specific intentions of the target user) through the text and / or language requests (queries), behavior processes, and expression feedbacks proposed by the target user during the interaction process, which are used to constitute the second learning information of the target user. And obtain the current learning state of the target user, the historical learning information of the target user, and the scene information where the target user is located, which constitute the first learning information of the target user for performing the subsequent steps S102 - S105.

[0079] Among them, it should be noted that the content and acquisition method of the learning needs proposed by this application for the target user are not limited, and can be selected or set according to the actual situation and empirical values. An optional implementation method is to obtain the learning needs proposed by the target user based on at least one round of conversation content between the large language model (LLM) and the target user.

[0080] In this implementation method, one possible source of the learning needs proposed by the target user (hereinafter referred to as ) is to directly obtain simple learning needs based on a single round of query between the LLM and the target user; another more complex learning need may be refined from multiple rounds of conversations between the LLM and the target user, which requires the refinement of the intentions of multiple historical rounds of conversations between the LLM and the target user. Another even more complex learning need may be to analyze by combining multi-modal behavior states. Here, taking this application as an example of analyzing and reasoning by combining multi-dimensional information such as the learning behavior, multiple rounds of interactions, and answering situations of the target user in the learning machine scenario, the learning needs of the target user can be accurately determined. The calculation formula is as follows:

[0081]

[0082] Among them, Router() means that different methods need to be self-consistently selected through the LLM (or other traditional small models such as pre-trained language representation models) by combining the actual effect and the scenario information where the target user is located for generation to improve the generation diversity and effect. Multi_feature represents multi-dimensional features, and the specific content is not limited. For example, Multi_feature can be expressed as the following example content:

[0083] {"User needs": "A set of exam papers suitable for me",

[0084] "Current portrait": "High fear of difficulty, weak foundation, tired of video learning...",

[0085] "Learning behavior": "Prefers objective questions, average focus duration within 30 minutes...",

[0086] "Historical interaction": "",

[0087] "Answering situation": "Accuracy rate of simple questions xx, accuracy rate of difficult questions xx, accuracy rate of different question types xx...",

[0088] }.

[0089] In addition, it should be noted that this application does not limit the content and acquisition method of the target user's current learning status, which can be selected or set according to the actual situation and empirical values. An optional implementation is that the target user's current learning status may include, but is not limited to, at least one of the target user's current expression, the text proposed by the target user (such as the target user entering the text "too difficult"), and / or the spoken language and behavioral expressions.

[0090] Moreover, this application does not limit the content and acquisition method of the target user's historical learning information, which can be selected or set according to the actual situation and empirical values. An optional implementation is that the target user's historical learning information may include, but is not limited to, at least one of the target user's historical behavior sequence, interaction sequence, answer content, process performance, and results.

[0091] Furthermore, this application does not limit the content and acquisition method of the scenario information where the target user is located, which can be selected or set according to the actual situation and empirical values. An optional implementation is that the scenario information where the target user is located may include, but is not limited to, at least one of the review scenario, preview scenario, exam preparation scenario, and vacation (such as summer and winter vacations) learning scenario. Among them, the learning focuses of each scenario are different. For example, the review scenario mainly focuses on the weak points and error-prone points of students, as well as the key points and difficult points of exams.

[0092] S102: Input the first learning information of the target user, combined with the first prompt instruction prompt, into the large language model to obtain the user portrait of the target user output by the model.

[0093] In this embodiment, it should be noted that in order to enable a more user-friendly and intelligent learning accompaniment process for the target user and improve the learning experience of the target user, after obtaining the first learning information of the target user (including but not limited to the current learning status of the target user (such as the current expression of the target user, the text and / or language and behavior expressions put forward), the historical learning information of the target user (such as historical behavior sequences, interaction sequences, answer contents, process performances, and results, etc.), and the scenario where the target user is located (such as review scenarios, preview scenarios, exam preparation scenarios, holiday learning scenarios, etc.) information) through step S101, further, based on the LLM technology, the unstructured information in these information that is difficult to process by traditional small models such as other pre-trained language representation models can be structured, and according to the full-scenario multi-modal learning features included in the processing results, a personalized user portrait of the target user can be constructed. Specifically, the current learning status, historical learning information, and scenario information of the target user included in the first learning information of the target user can be integrated into the prompt instruction prompt (here defined as the first prompt instruction) and input into the large language model (LLM) to obtain the personalized portrait of the target user output by the large language model. And, in order to enrich the portrait content of the target user, the traditional portrait features of the target user (such as basic attribute features, statistical attribute features, etc.) can also be combined to form the overall user portrait of the target user for performing the subsequent step S103.

[0094] Among them, the portrait content and classification of the target user in this application are not limited. For example, the portrait content of the target user can be divided into basic attributes, statistical attributes, and intrinsic attributes. Among them, the basic attributes refer to personal information such as the age, gender, geographical location, and occupation of the target user; the statistical attributes can be calculated from the teaching data according to the set statistical indicators. For example, the average correct rate of the target user's questions is statistically calculated from the static data, or the average viewing duration of the target user's online courses is statistically calculated according to the dynamic data; the intrinsic attributes refer to the hidden attributes that need to be mined based on multi-faceted learning condition data using methods such as expert knowledge and deep learning, such as the ability status and emotional status of students, etc. This requires special modeling and selection of data to obtain. For example, in this application, the following paradigm can be used when measuring the mastery degree and learning emotion of the target user for a certain knowledge point:

[0095] Mastery degree = KT(historical answer sequence, question characteristics, score)

[0096] Learning emotion = LM (historical interaction data, learning behavior characteristics)

[0097] Among them, LM represents the Language Model. Then, "learning sentiment = LM(historical interaction data, learning behavior characteristics)" means that the language model LM jointly analyzes the historical interaction data and learning behavior characteristics of the target user to generate or predict the learning sentiment of the target user. KT represents the Knowledge Tracing model. Then, "mastery = KT(historical answer sequence, question characteristics, score)" means that through the knowledge tracing model KT, combined with the historical answer sequence (such as historical answering patterns), question characteristics (such as question attributes), and scores (such as score details) of the target user, the real-time mastery of specific knowledge points is dynamically calculated.

[0098] In addition, to make up for the deficiency of simply reflecting the learning success of the target user (such as a student) by cognitive abilities such as intelligence and grades, this application proposes to include non-cognitive abilities such as perseverance and self-control, and non-intellectual factors such as learning motivation, self-awareness, and sound personality as an important part of the user portrait content. Such features cannot be effectively modeled by traditional small models such as pre-trained language representation models, such as being unable to understand user behavior, answering results, process emotions, interaction data, etc.

[0099] Thus, an optional implementation method is that after obtaining the unstructured information such as the historical learning information of the target user (such as historical behavior sequences, interaction sequences, answering content, process performance, and results, etc.) within a past period of time, the historical learning information of the target user during this period can be further structured, and the obtained processing result is incorporated into the first prompt instruction prompt and input into the large language model to obtain the historical portrait of the target user output by the model.

[0100] In this implementation method, structuring the historical learning information of the target user means structuring it according to different specific tasks, and the specific implementation method is not restricted. For example, the historical learning information data of students (as an example of the target user) can be uniformly converted into the json format. Examples of converting each category of data into a list and string through the model are as follows:

[0101] {

[0102] "answer sequence":[(question 1, 0.8),(question 2, 1.0),...],

[0103] "process performance":[(question 1, normal),(question 2, frowning),...],

[0104] }。

[0105] On this basis, integrating the above processing results into the prompt instruction, an example input to the large language model can be: "Please act as a subject teacher with rich teaching experience and conduct a step-by-step analysis of the student's {"historical structured features"}, including the student's learning content, performance in answering questions, and emotional performance during the process... etc., to form relevant portrait features, which include'student initiative, sense of purpose, current emotion, depth of usual thinking...'."

[0106] It can be understood that in this implementation method, the historical learning information content of the target user, such as the historical behavior sequence, interaction sequence, answer content, process performance, and results, is structured, and then the natural language understanding ability and thinking logic reasoning ability of the LLM itself are used for multi-link chain of thought (COT) reasoning analysis. For example, the learning motivation is analyzed by the LLM through the participation frequency, completion status, and the degree of seriousness of the content completed during the answer process of the student (as an example of the target user) (such as the richness of the notes after optical character recognition (OCR) compared to the standard answers, the process attitude, the clarity of the answer idea, the depth of thinking, etc.) and various learning behaviors after the answer result. After giving the expert constraint framework, let the LLM perform traceability reasoning analysis according to each dimension and data to judge whether the student (as an example of the target user) has strong initiative and sense of purpose.

[0107] In addition, another optional implementation method is that after obtaining the historical portrait of the target user (such as the historical portrait at the previous moment), the current learning state of the target user (such as the current expression of the target user, the text and / or language and behavior expressions put forward), and the scene information where the target user is located (such as review scene, preview scene, exam preparation scene, holiday learning scene, etc.), further integrate the current learning state of the target user at the current moment, the historical portrait at the previous moment, and the scene information where it is located into the updated first prompt instruction prompt, and input it into the large language model to obtain the instant portrait of the target user output by the model.

[0108] In this implementation method, for the immediate needs and emotional states of the target user (such as students) during the learning process, such as the student appearing tired and having a poor emotional state, etc., these immediate portrait captures are also crucial for accurately determining the accompanying learning strategy for the target user in the subsequent steps. Therefore, this application is based on the joint application of the LLM and related plugins, and represents the state of the target user (such as students) at each moment ( ), the historical portrait at the previous moment ( ), and the scene information where the target user is located ( ), incorporate a prompt instruction (denoted as P here) as the input to the LLM, so as to form a real-time generation of the target user's portrait through the natural language understanding ability and thinking logic reasoning ability inherent in the LLM, and obtain the target user's instant portrait (denoted as ), and the specific calculation formula is as follows:

[0109]

[0110] Among them, the purpose of the prompt instruction (P) is to generate the user portrait at a certain moment, and its main function is to incorporate relevant scenarios and portrait features into the prompt words with a thinking chain according to hyperparameters. An example can be: "Please act as a teacher with rich teaching experience, and by combining the current learning {status}, the student's {historical portrait features}, and the {scenario information} that the current student is learning, gradually analyze and combine to obtain the latest portrait features of the student, and generate features for updating the existing portrait dimensions. If new requirements are involved, they need to be condensed into

current requirements

[0111] while refers to a structured portrait, which updates some feature values of the existing historical portrait. Only the changed features are shown in the following examples. The changed feature values will affect the subsequent recommendation and dialogue strategies. For example, if a certain student skips difficult questions multiple times at a certain moment, then the example can be:

[0112] {

[0113] ……

[0114] "Fear of difficulty": low --> high,

[0115] "Abnormal behavior": frequent skipping of questions,

[0116] ……

[0117] }.

[0118] Furthermore, after obtaining the historical portrait of the target user and the instant portrait of the target user, the overall user portrait of the target user (denoted as U this time) can be constructed by combining the two with the target user's traditional portrait features (such as basic attribute features, statistical attribute features, etc.).

[0119] On this basis, by executing the subsequent step S103, the LLM can be used to analyze the learning needs of the target user (such as the decomposition of user intentions), and determine to call relevant agents to meet the real-time needs of the target user and adjust the learning strategy.

[0120] S103: Use the user profile of the target user, the second learning information of the target user, and the agent information, combined with the second prompt instruction "prompt", and input them into the large language model to obtain the learning strategy for the target user output by the model.

[0121] In this embodiment, after obtaining the user profile of the target user through step S102, and the first learning information such as the scene information (such as review scene, preview scene, exam preparation scene, holiday learning scene, etc.) where the target user is located and the second learning information such as the learning needs (i.e., the specific intentions of the target user are disassembled) proposed by the target user through step S101, further, the user profile of the target user, the learning needs proposed by the target user, the scene information where the target user is located, and the agent information can be integrated into the prompt instruction "prompt" (here it is defined as the second prompt instruction), and input into the large language model (LLM) to obtain the personalized learning strategy for the target user output by the large language model, which is used to execute the subsequent step S104.

[0122] Specifically, in order to be able to use the LLM to make decisions on the explicit / implicit needs shown by the target user, combined with the user profile and scene information, a variety of different scenarios, different needs, different implicit features are first constructed, and the strategy decision data brought by different profiles are combined to pre-train the LLM, so as to meet the actual needs of the target user in the real scene. And during the training process of the LLM, the descriptions and functions of different tools will be combined with the prompt for training. In this way, when the model combines the specific scene and user intention, it can effectively decide which appropriate tool to apply. The strategy decision can include but is not limited to tool invocation, learning plan generation, and evaluation feedback, etc.

[0123] In this way, when using the LLM to generate the learning strategy for the target user, the user profile (such as ∈U (including the instant profile + historical profile)) of the target user (such as a student), the learning needs ( ∈Q, can be absent) proposed by the target user, the scene information ( ∈S) where the target user is located, and the agent information (represented as R this time) can be integrated into the prompt instruction and used as the input of the LLM, so as to form the generation of the learning strategy for the target user at the t-th moment (represented as ) through the natural language understanding ability and thinking logic reasoning ability possessed by the LLM itself. The specific calculation formula is as follows:

[0124]

[0125] Among them, the prompt instruction adds the agent information (R) and the characteristics of the input parameters, and the output of the LLM is to form a specific learning strategy. Specifically, the LLM is required to generate a learning strategy based on the portrait of the target user, the application scenario and the real-time needs, combined with the input parameters and descriptions of the agent. For example, assuming that the target user has a high fear of difficulties and frequently skips questions, the strategy generated by the LLM after analysis may be: "Recommend objective questions with lower difficulty at the weak points", and the example content of R can be as follows:

[0126] R:{

[0127] R1: {"Question Paper Assistant": "Can build question papers and question packs according to the needs such as the scope and difficulty of questions specified by the user", "Input Parameters": [grade, book_code, error_topic,...]}

[0128] R2: {"Question Answering Teacher": "..."}

[0129] }。

[0130] S104: According to the learning strategy, the large language model is used to collaboratively call the agent to achieve accompanying learning for the target user.

[0131] In this embodiment, after obtaining the learning strategy for the target user through step S103, further, according to this learning strategy, the large language model can be used to call each agent to implement the main process of accompanying learning for the target user. And in the process of implementing the refined process of accompanying learning, through the mutual calls between each agent, the collaborative call of each agent is realized, so as to complete the entire process of accompanying learning for the target user.

[0132] For example: As Figure 2 shown, assume that the target user (such as a student) said a sentence "The final exam is coming soon. It would be great if I could have some special training" to the learning machine (deployed with the above-mentioned LLM) during a certain application exercise (such as Figure 2 the query in), at this time, through the LLM, it can quickly analyze that the intention of the target user (such as a student) is: improving scores before the exam, and the key elements of this intention are as Figure 2As shown by "bot: Final Exam, Score Improvement", afterwards, relevant learning profiles of the generated target users (such as students) can be called, and corresponding weak learning resources matching the target users (such as students) can be identified. Moreover, the learning benefits of these learning resources are calculated through relevant learning benefit models, and the specific calculation process is not limited. Then, in combination with the invocation of intelligent agents such as test paper generation and grading, the construction of the entire exercise package / volume is completed. After the target user (such as a student) finishes learning, the intelligent agent AI teacher and AI comment generation can be automatically invoked based on the reasons for the target user's (such as a student's) mistakes, realizing targeted learning and explanation of the questions that the target user (such as a student) does not understand and planning subsequent practice plans, etc. In this way, through an inadvertent sentence feedback from the target user (such as a student), targeted pre-exam practice and perfect guidance opinions can be completed. Among them, the invocation of the above-mentioned various intelligent agents or tools is completely triggered by the LLM generating instruction parameters or a combination mode of partial fixed workflows to ensure the accuracy of scheduling.

[0133] It should be noted that in the above scheduling process, the LLM participating in the scheduling is mainly responsible for connecting the main processes, while each intelligent agent (such as Figure 2 "test paper generation", "grading", "AI teacher" in Figure 2 itself can be scheduled with each other according to the environmental feedback. For example,

[0134] after the "AI teacher" finishes explaining the test questions to the target user (such as a student), it can schedule the "test paper generation" intelligent agent to generate grouped practice questions of the same type. In this way, the main process LLM makes decisions based on the user's intentions and profiles, and the internal operation processes of each intelligent agent in the detailed process can be coordinated and invoked, thus improving the decision-making efficiency.

[0134] On this basis, in order to further improve the companion learning effect for the target user and further enhance the learning experience of the target user, an optional implementation method is that during the process of realizing the companion learning for the target user by coordinating the invocation of each intelligent agent through the LLM, according to the feedback information of the target user, the user profile of the target user can be updated using the process self-reflection mechanism. Then, using the updated user profile of the target user, the learning strategy of the target user can be updated, and using the updated learning strategy of the target user, the intelligent agents can be coordinated and invoked again through the large language model to realize the intelligent companion learning for the target user until the learning needs put forward by the target user are met. Thus, the companion learning effect for the target user is effectively improved, and further, the learning experience of the target user is enhanced.

[0135] Specifically, in order to enable the accompanying learning system deployed with the LLM to analyze and sensitively understand the learning situation of the target user (such as a student) more like a human and timely give the correct learning strategy, the LLM needs to continuously reflect during the reasoning process, and adjust in real-time in combination with the individual portrait of the target user (such as a student). For example, after the target user (such as a student) receives the output result of the corresponding intelligent agent, what is the corresponding state? Whether it is timely and correctly feedback to the intelligent agent / LLM, etc., so as to enter a new round of update of the learning strategy, realizing a process of real-time monitoring, self-learning and reasoning error correction.

[0136] To achieve the above process and improve the accompanying learning effect, the present application proposes a "multi-layer self-reflective reasoning chain" based on the LLM, as Figure 3 shown, to increase the comparison (comparison with the original portrait, intention, context, feedback facts, etc.) and thinking depth in the reasoning process of the LLM. Specifically, it can include but is not limited to four key steps, namely: (1) Relevance between the recommended content and the intention, (2) Analysis of the real participation degree of the content, interaction & behavior analysis during the learning process, (3) Analysis of the learning results, (4) Update (such as adding / modifying) the user portrait. Among them, first, the LLM needs to analyze and reason each step in turn, and then reverse-infer and verify the results generated in each process in the next step. If it is reasonable, then update (such as adding / modifying) the user portrait in the last step. If the reverse inference is unreasonable, it is necessary to further trigger multi-round interaction confirmation or temporarily shelve it for verification. After the portrait of the target user is updated, a new learning strategy ( ) is triggered to ensure the real-time accompanying learning effect, thus greatly promoting the correctness of the analysis process and reducing the hallucination of the content generated by the LLM.

[0137] Here, it should be noted that generally the LLM base itself may not be able to complete the above step analysis well according to the prompt. Therefore, the present application separately constructs the above-mentioned thinking reasoning chain data, and then performs supervised fine-tuning training on the LLM. And, different from the traditional LLM decoding process that only outputs the topN to obtain the best result, during the application (reasoning) process of the LLM, the present application requires the LLM to be able to generate multiple results, and then score each step through the reward-model (scoring model) trained by the supervised task, so as to screen each step during the decoding process. The judgment of the best step comes from the feedback of the final user's learning achievement.

[0138] On this basis, an optional implementation method is that during the accompanying learning process, according to the feedback information of the target user, the implementation process of updating the user portrait of the target user by using the process self-reflection mechanism can specifically include the following steps A-D:

[0139] Step A: During the learning process, determine the target learning resources to be recommended based on the learning strategy; and determine whether the relevance of the target learning resources to the target user's learning needs and user profile meets the preset requirements.

[0140] In this implementation, after LLM outputs the learning strategy for the target user, it can further determine the learning resources to be recommended (this time it is defined as the target learning resource) according to the learning strategy; and judge whether the relevance between the target learning resource and the learning needs and user profile of the target user meets the preset requirements (the specific content is not limited and can be set according to the actual situation and experience value), such as Figure 3 As shown, to confirm self-consistency, if yes, proceed to the subsequent step B; if not, proceed to the subsequent step C.

[0141] Step B: If so, the target learning resources are recommended to the target user, and real-time interaction is performed with the target user based on the target user's learning participation, interaction during participation, and various learning behaviors and learning attitudes, so as to recommend the adjusted target learning resources to the target user and update the relevant content in the target user's user profile based on the target user's feedback information.

[0142] If it is determined through step A that the relevance of the target learning resource to the target user's learning needs and user profile meets the preset requirements (i.e., the content of the target learning resource meets the target user's learning intention), the target learning resource can be further recommended (output) to the target user, such as Figure 3 As shown, the learning participation of the target user is monitored, and the interaction and various learning behaviors and learning attitudes of the target user during the participation process are observed, such as low-frequency participation, exit from the learning process, poor performance during the process (such as expression of aversion to learning), etc., and real-time interaction is carried out with the target user based on this information, so as to further evaluate whether the content quality of the target learning resource meets the needs of the target user, whether the content is too difficult, the amount of content, and changes in the student's learning status, etc., based on the feedback information of the target user, so as to adjust the learning strategy and the target learning resources it contains, and recommend the adjusted target learning resources to the target user to perform the subsequent step D. For example, if the target user (such as a student) encounters a difficult problem and makes no progress, the relevant AI teacher can be called for auxiliary learning; or, if the target user (such as a student) is in a bad mood for learning, the relevant mood relief agent can be called for auxiliary learning. At the same time, the relevant content in the user portrait of the target user can also be updated.

[0143] Step C: If not, confirm the adjusted target learning resources through secondary interaction with the target user; recommend the adjusted target learning resources to the target user, and update the relevant content in the user profile of the target user.

[0144] If it is determined through step A that the relevance between the target learning resource and the learning needs and user profile of the target user does not meet the preset requirements (i.e., the content of the target learning resource does not conform to the learning intention of the target user), then this recommendation can be abandoned, and a secondary interaction can be carried out with the target user to confirm the adjusted target learning resource, and the adjusted target learning resource can be recommended to the target user for subsequent step D. At the same time, relevant content in the user profile of the target user can also be updated.

[0145] Step D: Analyze the learning results of the target user for the target learning resource and / or the adjusted target learning resource, and update the user profile of the target user according to the analysis results.

[0146] After recommending the target learning resource and / or the adjusted target learning resource to the target user through step B or C, further analyze the learning results of the target user for the target learning resource and / or the adjusted target learning resource, and update the user profile of the target user according to the analysis results.

[0147] Among them, the specific content of the analysis results in this application is not limited, and may include but is not limited to the insufficient completion degree of the learning content of the target user for the target learning resource and / or the target learning resource, and the error cause analysis results of wrong questions, etc.

[0148] Specifically, after the target user completes learning the target learning resource and / or the adjusted target learning resource, by evaluating learning achievements, such as insufficient completion degree of learning content, error cause analysis of wrong questions, etc., and interacting with the target user, portrait elements such as the weak points of the target user (such as students) are updated.

[0149] In this way, through the above reflection process, the accuracy of the recommendation of learning resources for the target user during the accompanying learning process can be improved, and problems can be discovered in a timely manner to form a reliable response, thereby improving the learning experience of the target user.

[0150] For example: Suppose the query proposed by the target user (such as a student) is: "Help me review the recent courses."

[0151] The content of the target learning resource is: ["Concept Explanation of Quadratic Equations with One Unknown.mp4", "Comprehensive Application of Quadratic Equations with One Unknown.mp4", "Exercises on the Concept of Quadratic Equations with One Unknown.txt"].

[0152] The user profile of the target user (such as a student): [Average student, high fear of difficulty, question - brushing type, lack of learning goal sense, current learning progress, textbook version, historical weak points...].

[0153] The LLM first analyzes the recent courses of the target user (such as a student). Through the learning progress and textbook version in the portrait, it is determined that it is the chapter of "Quadratic Equations with One Unknown", but the specific section cannot be locked, triggering a second interaction with the target user (such as a student): "Can you elaborate on which specific section under the quadratic equation with one unknown?"

[0154] The query that the target user (such as a student) puts forward again can be: "Basic concepts."

[0155] The adjusted target learning resource launched by the LLM can be: "Explanation of the Concepts of Quadratic Equations with One Unknown.mp4".

[0156] However, the target user (such as a student) quickly skips over during the learning process. This abnormal behavior triggers thinking about the portrait. The LLM discovers that the target user (such as a student) is of the question - brushing type, and further recommends the adjusted target learning resource as: "Exercises on the Concepts of Quadratic Equations with One Unknown", and recommends "Exercises on the Concepts of Quadratic Equations with One Unknown" with the difficulty gradually escalating from medium difficulty.

[0157] At the same time, use the answer results of the target user (such as a student) to update the portrait of the target user (such as a student). Whether "the concepts of quadratic equations with one unknown" is a continuous weak point, the time spent in the learning process, and the emotional analysis during the learning process are used to optimize the decision - making of learning strategies, such as the current learning state and recommendation satisfaction of the target user (such as a student) in the portrait.

[0158] In this way, by performing the above steps S101 - S104, for the first time, the LLM structures the unstructured information in multi - dimensional features such as the real - time dialogue, behavior, expression, and feedback of the target user, which is difficult to handle by traditional small models such as other pre - trained language representation models. According to the full - scene multi - modal learning features included in the processing results, a personalized user portrait of the target user is constructed. Then, through single / multi - round interaction, behavior process, expression feedback, etc., the learning intention of the target user is disassembled in real - time, and based on the roles required for accompanying learning, a decision on learning strategies is obtained to guide the coordinated invocation of multiple intelligent agents for multiple types of auxiliary tools, and a multi - intelligent - agent collaborative accompanying learning system with the LLM as the core decision - making brain is built to fully analyze the learning demands, learning states, and feelings of the target user (such as a student) throughout the process, and timely call and adjust learning strategies. Especially to improve the accompanying learning efficiency to achieve a human - like accompanying effect, this application introduces a process reflection mechanism during the accompanying learning process to promote the accurate construction of the target user portrait by the LLM, thereby more efficiently improving the timeliness and accuracy of the entire accompanying learning process. It realizes multi - functional intelligent accompanying learning, not only improving the educational quality of the target user, but also enhancing the fun of the target user during the learning process.

[0159] In summary, for the intelligent learning companion method provided in this embodiment, first, the first learning information and the second learning information of the target user are obtained; then, using the first learning information of the target user and combining it with the first prompt instruction "prompt", it is input into the LLM to obtain the user profile of the target user output by the model; next, using the user profile of the target user, the second learning information of the target user, and the agent information, and combining it with the second prompt instruction "prompt", it is input into the LLM to obtain the learning strategy for the target user output by the model; then, according to the learning strategy, the LLM is used to make a collaborative call to the agent to achieve the learning companion for the target user.

[0160] It can be seen that since this application first accurately constructs the full-scenario profile of the target user based on the LLM using the first learning information of the target user (such as the target user's historical learning information, current learning status, and the information of the scene where the user is located, etc.), and then determines the learning strategy based on the full-scenario profile of the target user, the second learning information of the target user (such as the learning requirements proposed by the target user, etc.), and the agent information, and according to this learning strategy, through the collaborative call of the agent and the process self-reflection mechanism, the intelligent learning companion for the target user is realized, effectively improving the timeliness and accuracy of the entire learning companion process, thereby improving the learning companion effect for the target user, and further enhancing the learning experience of the target user.

[0161] Second Embodiment

[0162] This embodiment will introduce an intelligent learning companion device. For related content, please refer to the above method embodiment.

[0163] See Figure 4 , which is a schematic diagram of the composition of an intelligent learning companion device provided in this embodiment. The device 400 includes:

[0164] An acquisition unit 401, configured to acquire the first learning information and the second learning information of the target user;

[0165] A first input unit 402, configured to use the first learning information of the target user and combine it with the first prompt instruction "prompt", and input it into the large language model to obtain the user profile of the target user output by the model;

[0166] A second input unit 403, configured to use the user profile of the target user, the second learning information of the target user, and the agent information, and combine it with the second prompt instruction "prompt", and input it into the large language model to obtain the learning strategy for the target user output by the model;

[0167] A learning companion unit 404, configured to make a collaborative call to the agent through the large language model according to the learning strategy to achieve the learning companion for the target user.

[0168] In one implementation of this embodiment, the first learning information of the target user includes the current learning status of the target user, the historical learning information of the target user, and the scenario information where the target user is located; the first input unit 402 includes:

[0169] A first input subunit, configured to structurally process the historical learning information of the target user, and input the obtained processing result combined with the first prompt instruction prompt into the large language model to obtain the historical portrait of the target user output by the model;

[0170] A second input subunit, configured to input the current learning status of the target user, the historical portrait of the target user, and the scenario information where the target user is located, combined with the updated first prompt instruction prompt, into the large language model to obtain the instant portrait of the target user output by the model;

[0171] A construction subunit, configured to construct the overall user portrait of the target user by using the historical portrait and the instant portrait of the target user.

[0172] In one implementation of this embodiment, the second learning information of the target user includes the learning requirements proposed by the target user; the learning requirements proposed by the target user are determined according to at least one round of conversation content between the large language model and the target user.

[0173] In one implementation of this embodiment, the second input unit 403 is specifically configured to:

[0174] Input the user portrait of the target user, the learning requirements proposed by the target user, the scenario information where the target user is located, and the agent information, combined with the second prompt instruction prompt, into the large language model to obtain the learning strategy for the target user output by the model.

[0175] In one implementation of this embodiment, the accompanying learning unit 404 is specifically configured to:

[0176] According to the learning strategy, call each agent through the large language model to implement the main process of accompanying learning for the target user; and in the process of implementing the refined process of accompanying learning, through the mutual call between each agent, realize the collaborative call of each agent, and complete the entire process of accompanying learning for the target user.

[0177] In one implementation of this embodiment, the device further includes:

[0178] An update unit, configured to update the user profile of the target user by using a process self-reflection mechanism according to the feedback information of the target user during the accompany learning process;

[0179] An invocation unit, configured to update the learning strategy of the target user by using the updated user profile of the target user, and then use the updated learning strategy to invoke the intelligent agent again through the large language model to achieve intelligent accompany learning for the target user until the learning requirements proposed by the target user are met.

[0180] In an implementation manner of this embodiment, the update unit includes:

[0181] A judgment subunit, configured to determine the target learning resources to be recommended according to the learning strategy during the accompany learning process; and judge whether the relevance of the target learning resources to the learning requirements and user profile of the target user meets the preset requirements;

[0182] A first update subunit, configured to, if it is determined that the relevance of the target learning resources to the learning requirements and user profile of the target user meets the preset requirements, recommend the target learning resources to the target user, and conduct real-time interaction with the target user according to the learning participation degree, interaction during the participation process, various learning behaviors and learning attitudes of the target user, so as to recommend the adjusted target learning resources to the target user according to the feedback information of the target user, and update the relevant content in the user profile of the target user;

[0183] A second update subunit, configured to, if it is determined that the relevance of the target learning resources to the learning requirements and user profile of the target user does not meet the preset requirements, confirm the adjusted target learning resources through secondary interaction with the target user; and recommend the adjusted target learning resources to the target user, and update the relevant content in the user profile of the target user;

[0184] A third update subunit, configured to analyze the learning results of the target user for the target learning resources and / or the adjusted target learning resources, and update the user profile of the target user according to the analysis results.

[0185] In an implementation manner of this embodiment, the analysis results include the insufficient completion degree of the learning content of the target user for the target learning resources and / or the target learning resources and the error cause analysis results of the wrong questions.

[0186] In one implementation of this embodiment, the current learning state of the target user includes at least one of the current expression of the target user, the text proposed by the target user, and / or the verbal and behavioral expressions made; the historical learning information of the target user includes at least one of the historical behavior sequence, interaction sequence, answer content, process performance, and result of the target user; the scenario information where the target user is located includes at least one of a review scenario, a preview scenario, a preparation scenario for an exam, and a holiday learning scenario.

[0187] Furthermore, an intelligent learning companion device is provided in an embodiment of the present application, including: a processor, a memory, and a system bus;

[0188] The processor and the memory are connected through the system bus;

[0189] The memory is used to store one or more programs, and the one or more programs include instructions that, when executed by the processor, cause the processor to execute any implementation method of the above intelligent learning companion method.

[0190] Furthermore, a computer-readable storage medium is provided in an embodiment of the present application. Instructions are stored in the computer-readable storage medium, and when the instructions are run on a terminal device, the terminal device is caused to execute any implementation method of the above intelligent learning companion method.

[0191] Furthermore, a computer program product is provided in an embodiment of the present application. When the computer program product runs on a terminal device, the terminal device is caused to execute any implementation method of the above intelligent learning companion method.

[0192] From the description of the above embodiments, those skilled in the art can clearly understand that all or part of the steps in the above embodiment methods can be implemented by means of software plus a necessary general hardware platform. Based on this understanding, the technical solution of the present application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. The computer software product can be stored in a storage medium, such as ROM / RAM, magnetic disk, optical disc, etc., and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network communication device such as a media gateway, etc.) to execute the methods described in each embodiment or some parts of the embodiments of the present application.

[0193] It should be noted that the various embodiments in this specification are described in a progressive manner. Each embodiment focuses on the differences from other embodiments. The same or similar parts among the various embodiments can be referred to each other. For the devices disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple, and the relevant parts can be referred to the description of the method part.

[0194] It should also be noted that, in this text, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device comprising a series of elements not only includes those elements, but also includes other elements not expressly listed, or elements inherent to such process, method, article or device. Without further limitation, an element defined by the statement "comprising an..." does not exclude the presence of additional identical elements in the process, method, article or device comprising the said element.

[0195] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be obvious to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. An intelligent learning companion method, characterized in that, Including: Obtain the first learning information and the second learning information of the target user; Use the first learning information of the target user, combined with the first prompt instruction prompt, and input it into the large language model to obtain the user profile of the target user output by the model; Use the user profile of the target user, the second learning information of the target user, and the agent information, combined with the second prompt instruction prompt, and input it into the large language model to obtain the learning strategy for the target user output by the model; According to the learning strategy, through the large language model, the agents are called collaboratively to achieve accompanying learning for the target user; The first learning information of the target user includes the current learning status of the target user, the historical learning information of the target user, and the scenario information where the target user is located; the step of using the first learning information of the target user, combined with the first prompt instruction prompt, and inputting it into the large language model to obtain the user profile of the target user output by the model includes: Structurally process the historical learning information of the target user, and use the obtained processing result combined with the first prompt instruction prompt, and input it into the large language model to obtain the historical profile of the target user output by the model; Use the current learning status of the target user, the historical profile of the target user, and the scenario information where the target user is located, combined with the updated first prompt instruction prompt, and input it into the large language model to obtain the instant profile of the target user output by the model; Use the historical profile and the instant profile of the target user to construct the overall user profile of the target user.

2. The method according to claim 1, characterized in that The second learning information of the target user includes the learning needs proposed by the target user; the learning needs proposed by the target user are determined according to at least one round of conversation content between the large language model and the target user.

3. The method according to claim 2, characterized in that, The step of using the user profile of the target user, the second learning information of the target user, and the agent information, combined with the second prompt instruction prompt, and inputting it into the large language model to obtain the learning strategy for the target user output by the model includes: Use the user profile of the target user, the learning needs proposed by the target user, the scenario information where the target user is located, and the agent information, combined with the second prompt instruction prompt, and input it into the large language model to obtain the learning strategy for the target user output by the model.

4. The method according to claim 1, characterized in that, The step of according to the learning strategy, through the large language model, the agents are called collaboratively to achieve accompanying learning for the target user includes: According to the learning strategy, through the large language model, each agent is called to achieve the main process of accompanying learning for the target user; and in the process of implementing the refined process of accompanying learning, through the mutual call between each agent, the collaborative call of each agent is realized, and the entire process of accompanying learning for the target user is completed.

5. The method according to any one of claims 1 to 4, characterized in that, The method further includes: During the process of accompanying learning, according to the feedback information of the target user, use the process self-reflection mechanism to update the user profile of the target user; Use the updated user profile of the target user to update the learning strategy of the target user, and use the updated learning strategy to cooperate with the large language model to call the agent again to achieve intelligent accompanying learning for the target user until the learning needs proposed by the target user are met.

6. The method according to claim 5, characterized in that, The process of updating the user profile of the target user by using the process self-reflection mechanism according to the feedback information of the target user during the accompanying learning process includes: During the accompanying learning process, determine the target learning resources to be recommended according to the learning strategy; and judge whether the relevance of the target learning resources to the learning needs and user profile of the target user meets the preset requirements; If so, recommend the target learning resources to the target user, and interact with the target user in real time according to the learning participation, interaction during the participation, various learning behaviors and learning attitudes of the target user, so as to recommend the adjusted target learning resources to the target user according to the feedback information of the target user, and update the relevant content in the user profile of the target user; If not, confirm the adjusted target learning resources through a second interaction with the target user; and recommend the adjusted target learning resources to the target user, and update the relevant content in the user profile of the target user; Analyze the learning results of the target user for the target learning resources and / or the adjusted target learning resources, and update the user profile of the target user according to the analysis results.

7. The method according to claim 6, wherein The analysis results include the insufficient completion degree of the learning content of the target user for the target learning resources and / or the target learning resources and the error cause analysis results of the wrong questions.

8. The method according to claim 1, characterized in that, The current learning state of the target user includes at least one of the current expression of the target user, the text proposed by the target user and / or the language and behavior expressions issued; The historical learning information of the target user includes at least one of the historical behavior sequence, interaction sequence, answer content, process performance and results of the target user; the scenario information where the target user is located includes at least one of the review scenario, preview scenario, exam preparation scenario, and holiday learning scenario.

9. An intelligent learning companion device, characterized in that, It includes: An acquisition unit for acquiring the first learning information and the second learning information of the target user; A first input unit for inputting the first learning information of the target user into the large language model in combination with the first prompt instruction prompt to obtain the user profile of the target user output by the model; A second input unit for inputting the user profile of the target user, the second learning information of the target user and the agent information into the large language model in combination with the second prompt instruction prompt to obtain the learning strategy for the target user output by the model; An accompanying learning unit for realizing the accompanying learning of the target user by cooperatively calling the agent through the large language model according to the learning strategy; The first learning information of the target user includes the current learning status of the target user, the historical learning information of the target user, and the scenario information where the target user is located; The first input unit includes: A first input subunit, configured to perform structured processing on the historical learning information of the target user, and input the obtained processing result combined with the first prompt instruction prompt into the large language model to obtain the historical portrait of the target user output by the model; A second input subunit, configured to input the current learning status of the target user, the historical portrait of the target user, and the scenario information where the target user is located, combined with the updated first prompt instruction prompt, into the large language model to obtain the instant portrait of the target user output by the model; A construction subunit, configured to construct the overall user portrait of the target user by using the historical portrait of the target user and the instant portrait of the target user.

10. An intelligent learning companion device, characterized in that, Including: A processor, a memory, and a system bus; The processor and the memory are connected through the system bus; The memory is used to store one or more programs, and the one or more programs include instructions that, when executed by the processor, cause the processor to execute the method according to any one of claims 1-8.

11. A computer-readable storage medium, characterized in that, Instructions are stored in the computer-readable storage medium, and when the instructions are run on the terminal device, the terminal device is caused to execute the method according to any one of claims 1-8.

Images