Generative man-machine fusion student portrait analysis method and system, terminal and medium

Abstract

The invention provides a generation type man-machine fusion student portrait analysis method and system, and the method comprises the steps: obtaining the multi-modal data of a student, carrying out the fusion of the multi-modal data, and extracting the features of a student portrait; performing clustering processing under the guidance of a generative language large model on the student portrait features to obtain multiple groups of student portrait clustering data and a portrait vector database thereof; aiming at the student portrait clustering data, analyzing structured and unstructured information in the student portrait clustering data by utilizing a generative large language model, and identifying index information and risk factor information required for portrait personalized analysis; and on the basis of the portrait vector database, generating personalized prediction and recommendation information of the to-be-predicted student by using the retrieval-enhanced generative large language model. According to the method, based on a large language model, fusion and vectorization embedding are carried out by using multi-modal information of students, meanwhile, summarization analysis is carried out on portraits in growth of the students, future development after graduation is predicted, and generative personalized prediction information is provided.

Description

Technical Field

[0001] The present invention relates to the field of digital portrait technology, and in particular to a method, system, terminal and medium for analyzing student portraits in a human-machine integrated manner under generative artificial intelligence. Background Art

[0002] With the development of mobile Internet and big data technology, the application of digital portrait technology in the field of education has received more and more attention. Student portraits use standardized behavior and content data provided by the data center, such as course grades, library borrowing, etc., and use big data analysis methods to quantify students' multi-dimensional characteristics such as regularity, effort, learning skills, social relationships, etc., reveal students' growth trajectories, and predict students' academic performance, employment tendencies, psychological conditions, etc. based on prediction models, so as to provide an important basis for schools to conduct personalized and precise education management and guidance for students. Therefore, student portraits have become an important target point in the current education field.

[0003] Existing student profiling technology can usually achieve the following functions:

[0004] -Student basic information (name, date of birth, mobile phone number, class, student number, etc.);

[0005] -Instructor information;

[0006] - Grades (GPA);

[0007] -Experience (e.g. four years of college, record college activities on a monthly or quarterly visual timeline);

[0008] - Comprehensive quality (ideology and politics, honors, social practice, volunteer service, student activities, award-winning experience, personal summary);

[0009] - Admission application;

[0010] - Admission information (interest in the major you are applying for, your understanding of the college, your English proficiency, and your understanding of scientific research);

[0011] - Graduation information (admission speech, graduation speech);

[0012] - Statistical information such as national enrollment ratio, etc.;

[0013] Most existing student portrait technologies are merely collections of various data on students and teachers. They are seriously inadequate in integrating student portrait information and do not have predictive analysis capabilities. They lack intelligent generation output from the perspective of students and management. Although existing student portrait technologies can partially meet the basic information queries of students and teachers, they are essentially unable to provide deeper insights into student data, such as summaries of the overall portrait of teachers and students, personalized recommendations for students' future graduation destinations, and predictions of future academic or work performance. Summary of the invention

[0014] In view of the above-mentioned deficiencies in the prior art, the present invention provides a student portrait analysis method, system, terminal and medium with human-machine fusion under generative artificial intelligence.

[0015] According to one aspect of the present invention, a generative human-machine fusion student portrait analysis method is provided, comprising:

[0016] Acquire multimodal data of students, fuse the multimodal data, and extract student portrait features;

[0017] Providing a generative large language model, performing clustering processing on the student portrait features under the guidance of the generative large language model, and obtaining multiple groups of student portrait clustering data and their portrait vector databases;

[0018] For each group of the student portrait cluster data, the generative large language model is used to analyze the student portrait cluster data to identify the indicator information and risk factor information required for the personalized analysis of the existing student portraits;

[0019] The basic data of the new student to be predicted is taken as input data. Based on the portrait vector data similar to the student to be predicted, the retrieval-enhanced generative large language model is used to perform personalized portrait prediction on the student to be predicted and generate corresponding customized recommendation information.

[0020] According to another aspect of the present invention, a generative human-machine fusion student portrait analysis system is provided, comprising:

[0021] A multimodal data fusion module, which is used to obtain student multimodal data, fuse the multimodal data, and extract student portrait features;

[0022] A feature clustering module, which is used to perform clustering processing on the student portrait features under the guidance of a generative large language model to obtain multiple groups of student portrait clustering data and their portrait vector databases;

[0023] An information extraction module, which is used to analyze the student portrait clustering data using a generative large language model for each group of the student portrait clustering data, and identify the indicator information and risk factor information required for the personalized analysis of the existing student portraits;

[0024] The result recommendation module is used to take the basic data of the new student to be predicted as input data, and based on the portrait vector data similar to the student to be predicted, use the retrieval-enhanced generative large language model to make personalized portrait predictions for the student to be predicted, and generate corresponding customized recommendation information.

[0025] According to a third aspect of the present invention, there is provided a computer terminal comprising a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, can be used to execute any one of the methods described above in the present invention, or to run the system described above in the present invention.

[0026] According to a fourth aspect of the present invention, there is provided a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, can be used to execute any of the methods described above in the present invention, or to run the system described above in the present invention.

[0027] Due to the adoption of the above technical solution, the present invention has at least one of the following beneficial effects compared with the prior art:

[0028] The generative human-machine fusion student portrait analysis method, system, terminal and medium provided by the present invention utilize the personalized prediction function driven by a large language model (a generative large language model and a generative large model based on retrieval enhancement) to generate customized predictive recommendation data by analyzing various data such as academic records, extracurricular activities, teacher-student interactions and personal preferences, so that students can have auxiliary reference for their academic path, course selection and employment strategy, thereby improving their academic performance and career achievements.

[0029] The generative human-machine fusion student portrait analysis method, system, terminal and medium provided by the present invention use the predictive analysis capability driven by a large language model to support students in early intervention of interference factors. By identifying early indicators of student success or risk factors for academic downturn, proactive intervention measures are taken accordingly, and targeted support and intervention are provided to help students stay on the right track and achieve their academic goals, thereby improving academic outcomes and enhancing learning achievements.

[0030] The generative human-machine fusion student portrait analysis method, system, terminal and medium provided by the present invention utilize the resource allocation and operational efficiency functions driven by the large language model to help optimize resource utilization by analyzing course selection patterns, student preferences and academic performance data, and make data-driven decisions based on student needs, such as assigning tutors for academic support services or which courses to offer. This resource optimization improves educational operational efficiency and ensures that resources are allocated in a way that maximizes student learning achievement.

[0031] The generative human-machine fusion student portrait analysis method, system, terminal and medium provided by the present invention utilize the auxiliary detection driven by a large language model and the elimination of bias in data and algorithms to promote transparency and fairness in decision-making, while protecting students' privacy and data security, and promoting fair and inclusive practices. By incorporating unstructured data such as moral and ideological considerations into student evaluation incentives, we ensure that policies are in the best interests of all students and create a more equitable and inclusive learning environment.

[0032] The generative human-machine fusion student portrait analysis method, system, terminal and medium provided by the present invention are driven by a large language model to achieve personalized learning experience, early intervention of interference factors, improvement of academic performance and participation, optimization of resource allocation, and fairness and inclusiveness practices. By combining the large language model algorithm framework generated by retrieval enhancement, enhanced educational results and richer learning experience are achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Other features, objects and advantages of the present invention will become more apparent from the detailed description of non-limiting embodiments made with reference to the following drawings:

[0034] Figure 1 This is a workflow diagram of a generative human-machine fusion student portrait analysis method in a preferred embodiment of the present invention.

[0035] Figure 2 This is a working block diagram of a generative human-machine fusion student portrait analysis method in a preferred embodiment of the present invention.

[0036] Figure 3 Schematic diagram of the student portrait analysis work framework based on generative human-machine fusion under a retrieval enhancement framework in a preferred embodiment of the present invention.

[0037] Figure 4 Schematic diagram of the component modules of a generative human-machine fusion student portrait analysis system in a preferred embodiment of the present invention. DETAILED DESCRIPTION

[0038] The following is a detailed description of the embodiments of the present invention: This embodiment is implemented on the premise of the technical solution of the present invention, and a detailed implementation method and a specific operation process are given. It should be pointed out that for ordinary technicians in this field, several variations and improvements can be made without departing from the concept of the present invention, which all belong to the protection scope of the present invention.

[0039] An embodiment of the present invention provides a generative human-machine fusion student portrait analysis method, which is driven by a large language model and presents the portrait characteristics of students in various aspects in human-machine fusion based on a retrieval enhancement generation algorithm driven by the large language model. It actively presents to students a visual display of the growth records of students' educational stages and future development predictions after graduation according to a timeline; by providing personalized information, it can promote early intervention, improve academic performance, enhance student participation, optimize resource allocation, and promote inclusiveness and fairness, thereby improving the education landscape.

[0040] Specifically, Figure 1 As shown, the generative human-machine fusion student portrait analysis method provided in this embodiment may include the following operations:

[0041] S1, obtain students’ multimodal data, fuse the multimodal data, and extract student portrait features;

[0042] S2, providing a generative large language model, performing clustering processing on student portrait features under the guidance of the generative large language model, and obtaining multiple groups of student portrait clustering data and their portrait vector databases;

[0043] S3, for each group of student portrait cluster data, use the generative large language model to analyze the student portrait cluster data and identify the indicator information and risk factor information required for personalized analysis of existing student portraits;

[0044] S4, taking the basic data of the new student to be predicted as input data, based on the portrait vector data similar to the student to be predicted, uses the retrieval-enhanced generative large language model to make personalized portrait predictions for the student to be predicted, and generates corresponding customized recommendation information.

[0045] In some preferred embodiments, the above method may further include the following operations:

[0046] Use generative large language models to automatically generate dialogues, achieve human-computer interaction fusion, and obtain automated dialogue communication and support information.

[0047] In some preferred implementations, the above S1, obtaining multimodal data of students, fusing the multimodal data, and extracting student portrait features, may further include the following operations:

[0048] S11, obtain student multimodal data, including: existing structured and unstructured information in the internal school teacher-student system, as well as unstructured information such as natural language descriptions, pictures, videos, etc. in external resources and information;

[0049] S12, providing a pre-trained Chinese language model, and using the Chinese language model to vectorize the unstructured information to obtain vector data;

[0050] S13, fuses the vector data and structured information, and extracts features from the fused data to obtain student portrait features.

[0051] In some preferred implementations, the above S12, the pre-trained Chinese large language model, is obtained by training an open source large language model using structured and unstructured information in the internal school teacher-student system.

[0052] In some preferred implementations, the above S2 provides a generative large language model, performs clustering processing on student portrait features under the guidance of the generative large language model, and obtains multiple groups of student portrait clustering data and their portrait vector databases, and can further include the following operations:

[0053] S21, providing a generative large language model, based on the student portrait features, using the prompt engineering of the generative large language model to guide the generation of different types of student portrait descriptions, and obtaining a portrait cluster description;

[0054] S22, use the embedding model of the generative large language model to encode the student multimodal data into vectors, calculate the similarity between each student and the portrait cluster description, and assign the students to portrait categories with high similarity (i.e., the similarity is higher than the set threshold), obtain multiple groups of student portrait cluster data containing structured information and unstructured information, and construct a portrait vector database.

[0055] In some preferred implementations, the above S3, for each group of student portrait cluster data, uses a generative large language model to analyze the student portrait cluster data to identify the indicator information and risk factor information required for personalized prediction of existing student portraits, and may further include the following operations:

[0056] S31, using the understanding and generation capabilities of the generative large language model to analyze the structured and unstructured information in the student portrait clustering data, extract the set features of different portrait clusters, and generate key indicator information and several risk factor information for personalized predictions of existing students and future students.

[0057] In some preferred implementations, the above S4, taking the basic data of the new student to be predicted as input data, based on the portrait vector data similar to the student to be predicted, using the retrieval-enhanced generative large language model to perform personalized portrait prediction on the student to be predicted, and generating corresponding customized recommendation information, may further include the following operations:

[0058] The newly added structured and unstructured information in the internal school teacher-student system is used as the basic data of new students to be predicted. Based on the portrait information of student groups similar to the basic data of students, the algorithm framework of the generative large language model based on retrieval enhancement is used. The records of similar students in the portrait vector database and the similar portrait clustering centers are used as context. The personalized risk factor information and indicator information analysis of the basic data of students is performed to obtain the portrait prediction results of personalized predictions for the future of new students, and the corresponding customized recommendation information is generated. The prediction of the students' future graduation and employment portraits is given, and the portrait analysis and prediction of the students to be predicted are completed, which effectively improves the overall student experience, and contributes to the students' academic success and personalized recommendations for future study abroad or employment.

[0059] In some preferred implementations, the above method uses the basic data of the new student to be predicted as input data, and based on the portrait vector data similar to the student to be predicted, uses the retrieval-enhanced generative large language model to perform personalized portrait prediction on the student to be predicted, and generates corresponding customized recommendation information, specifically including:

[0060] S41, using the generative large language model and combining it with the retrieval enhancement algorithm to establish a retrieval enhanced generative large language model;

[0061] S42, obtaining newly added structured and unstructured information in the internal school teacher-student system as basic data of new students to be predicted;

[0062] S43, the retrieval part of the retrieval-enhanced generative large language model uses student portrait clustering analysis to retrieve student information related to the student to be predicted in the portrait vector database (which may include student portrait feature data, such as academic performance, hobbies and interests, social activities and other school internal and external information, and / or key indicator information and several risk factor information), and obtains a portrait description of a similar student group;

[0063] S44, using the retrieval results of similar student group portrait descriptions obtained by the retrieval part of the generative big model as context information together with the basic data of the new student to be predicted as the input of the generation part of the generative big model, predicting the indicator information and risk factor information of the basic data of the new student to be predicted, thereby generating personalized analysis results of the new student portrait and generating corresponding customized recommendation information.

[0064] In some preferred implementations, the above S45, customized recommendation information, includes: key indicator analysis result information of the portrait, personalized portrait prediction recommendation information, and evaluation incentive information. It may also include: automated dialogue communication and support information.

[0065] The technical solution provided by the above embodiment of the present invention is further described in detail below in conjunction with a specific application example.

[0066] like Figure 2 As shown in the figure, it is a working architecture diagram of the student portrait analysis method of generative human-machine fusion involved in this specific application example, which mainly includes the following steps:

[0067] Step 1: Multimodal data fusion: Obtain students’ multimodal data, fuse the multimodal data, and extract student portrait features.

[0068] This step specifically includes:

[0069] Utilize the basic information of students in school or graduating, including structured and unstructured natural text data, and associate it with news reports about the student appearing on campus, including understanding of pictures and videos of the student appearing in the news reports. At the same time, automatically crawl the information about the study abroad or employment of graduated students to the school student portrait database system. Use the pre-trained Chinese large language model to vectorize the student information data and store it in the vector database. Furthermore, structured student information: such as name, date of birth, mobile phone, class, student number, tutor information, grades, GPA, etc.; unstructured information described in natural language: such as university activities, comprehensive quality, ideology and politics, honors, social practice, volunteer service, student activities, award-winning experience, graduation summary, etc. recorded during the four years of university experience.

[0070] Step 2: Automatic clustering of student portraits based on the generative large language model: cluster the student portrait features to obtain multiple groups of student portrait data.

[0071] This step specifically includes:

[0072] Using the prompt engineering of the generative large language model to cluster student portraits, first use the generative large language model to guide the generation of different types of student portrait descriptions. For example, the cluster description {cluster_description} in the algorithm framework 1 can be described as: "Outstanding academic achievements, strong research ability", "Active in community activities, outstanding leadership", "Strong innovation and entrepreneurship ability, rich practical experience", "Excellent volunteer service performance, strong sense of social responsibility", "All-round development, comprehensive balance of morality, intelligence, physical fitness, aesthetics and labor". Then use the embedding model of the generative large language model to encode the student's structured and unstructured information and student portrait descriptions into vectors. Finally, calculate the cosine similarity between each student and multiple portrait cluster descriptions generated under the guidance of the customizable generative large language model, and assign the student to the portrait category with the highest similarity. According to the student portrait features, use the pre-trained Chinese large language model to encode the student's structured and unstructured information descriptions into vectors. Then use the pre-trained model to calculate the similarity between each student and the portrait cluster description, and assign the student to the portrait category with high similarity. Identify groups of students with similar characteristics, interests or needs in the context of student profiles. Understanding these groups allows universities to better tailor their support services, academic programs, and interventions to meet the diverse needs of their student populations.

[0073] Step 3: Predictive analysis and early intervention: For each set of student portrait data, use a generative large language model to analyze the structured and unstructured information in the student portrait data to identify the indicator information and risk factor information needed for personalized prediction of existing students.

[0074] This step specifically includes:

[0075] Through the prompt engineering in the big model, we analyze the structured (such as grades, etc.) and unstructured data (such as comprehensive quality, etc.) in a large number of student portraits, identify early indicators of student success or risk factors of academic deficiencies, and predict future results, such as academic success, graduation rate, employment rate, study abroad rate, etc. Through these pattern and trend identification and insight generation, university administrators can intervene early to support and help students who may fall behind.

[0076] For obtaining sentiment analysis information of the portrait, it includes: using the unstructured data in the portrait system, automatically analyzing the key indicators and risk factors of the student portrait based on the generative large language model prompt engineering, so as to understand the emotions and opinions of the students in their growth. Sentiment analysis can provide valuable insights into students' experience, satisfaction level and areas for improvement in the university environment. This information can guide strategic decisions related to curriculum development, campus policies and student support services, and provide relevant explanations for the model results, that is, using the large language model to give explanations for the key indicators or predictions in the student portrait, such as why a certain classmate is predicted to drop out of school, because the lack of extracurricular activities may lead to social loneliness, etc.

[0077] By using the understanding and generation capabilities of the generative large language model, we can summarize the characteristics of different portrait clusters, such as successful graduates and dropouts, and generate key indicators and risk factors for personalized prediction of future students. This can be achieved using a prompt engineering algorithm program example similar to the following:

[0078] #Algorithm framework 1:

[0079] #Customize cluster description

[0080] cluster_description="""

[0081] A. Outstanding academic achievements and strong research capabilities: Students in this group generally perform well in their respective professional fields, demonstrating a solid academic foundation and strong research capabilities.

[0082] B. Active in club activities and outstanding leadership: Students actively participate in various club activities, demonstrating outstanding organizational and leadership skills, which provides a good platform for the cultivation of teamwork and social skills.

[0083] C. Strong innovation and entrepreneurship capabilities and rich practical experience: Students in this group generally have strong innovation awareness and entrepreneurial capabilities, often participate in various practical activities, and have accumulated rich real project experience.

[0084] D. Excellent volunteer performance and strong sense of social responsibility: Students actively participated in volunteer activities, demonstrated a strong sense of social responsibility, and were committed to serving the society and helping others.

[0085] E. All-round development, balanced morality, intelligence, physical fitness, aesthetics and labor: students develop in a balanced way in morality, intelligence, physical fitness, aesthetics and labor, showing the results of comprehensive quality education and being able to adapt to the diverse needs of society.

[0086] """

[0087] #Build Tips

[0088] prompt=f"""Based on the following student profile cluster descriptions, identify the common characteristics, potential success indicators, and risk factors of each profile cluster group.

[0089] Particular attention is paid to factors that may affect academic success across the cohort, including indicators of risk for dropout and outstanding graduation.

[0090] Cluster description: {cluster_description}

[0091] Student Group Information:

[0092] """

[0093] for student in students[:n]:#Limit to the first n students to control input length

[0094] prompt+=f"""

[0095] Student: {student['name']}

[0096] Major: {student['major']}

[0097] GPA: {student['gpa']}

[0098] Activities: {student['activities']}

[0099] Achievements: {student['achievements']}

[0100] Others: {student['others']}

[0101] """

[0102] prompt+="""

[0103] You are an experienced education assessment expert who specializes in analyzing the comprehensive performance and characteristics of college students. Here are two tasks

[0104] 1) Conduct cluster analysis of student portraits and give specific cluster labels based on each student’s information.

[0105] 2) For each student group in the profile clustering, analyze the results of the profile clustering. Please output the analysis results in JSON format, including the following keys: 'common_characteristics', 'success_indicators', 'risk_factors'.

[0106] Each key corresponds to a list, and each item in the list should have a 'factor' and an 'explanation'.

[0107] Output format example:

[0108]

[0109]

[0110] Step 4: Take the basic data of the student to be predicted (i.e. the basic data of the new student) as input data, and use the retrieval-enhanced generative large language model to make personalized predictions on the portraits of the students to be predicted based on the portrait vector data similar to the students to be predicted, and generate corresponding customized recommendation information; including: portrait sentiment analysis information, personalized recommendation information, automated dialogue communication, and evaluation incentive information, etc.

[0111] This step specifically includes:

[0112] To obtain personalized recommendation information, it includes: using the retrieval-enhanced generative large language model to drive the recommendation system, providing highly personalized suggestions for courses, extracurricular activities, resources and support services. By analyzing students' past learning history, interests, preferences, and even conversation data with counselors, using the retrieval-enhanced generative large language model algorithm framework, customized recommendations that meet the unique needs and goals of each student can be generated under the generative large language model. Improving the overall student experience will help their academic success and provide personalized recommendations for future study abroad or employment.

[0113] Using the algorithm framework of the retrieval-enhanced generative large language model, based on the information of similar students in the group clustering portrait, we can predict the future graduation and employment of new students. This can be achieved using a program example similar to the following:

[0114] #Algorithm framework 2:

[0115] #Generate predictions for students' future portraits

[0116] prompt=f"""Predict the student's future graduation and employment status based on the following information:

[0117] Query student information:

[0118] Major:{query_student['major']}

[0119] GPA:{query_student['gpa']}

[0120] Activities: {query_student['activities']}

[0121] Achievements:{query_student['achievements']}

[0122] Information about similar student groups in portrait clustering:

[0123] """

[0124] for i,student in enumerate(similar_students,1):

[0125] prompt+=f"""

[0126] Student {i}:

[0127] Major: {student['major']}

[0128] GPA:{student['gpa']}

[0129] Activities: {student['activities']}

[0130] Achievements: {student['achievements']}

[0131] Graduation status: {student['graduation_status']}

[0132] Employment status: {student['employment_status']}

[0133] Others: {student[others]}

[0134] """

[0135] prompt+="""

[0136] Based on the above information, please predict the possible graduation status and employment prospects of the queried students. Provide detailed analysis and reasons, including:

[0137] 1. Expected graduation status (whether to graduate on time, whether to be awarded outstanding graduate, etc.)

[0138] 2. Possible career directions and prospects

[0139] 3. Key factors affecting students’ future development

[0140] 4. Advice for students (how to improve graduation and employment success rates)

[0141] Please provide a detailed analysis:

[0142] """

[0143] .

[0144] like Figure 3 As shown, the algorithm framework of the generative large language model student portrait algorithm based on human-machine fusion under the retrieval enhancement framework is shown in the figure: the algorithm architecture of the generative large language model based on retrieval enhancement includes: an efficient retrieval part and a generation part. The efficient retrieval part quickly retrieves similar student group portrait information related to the user request in the vector database through student portrait clustering analysis, such as basic student data, such as academic performance, interests and hobbies, social activities and other internal and external information of the school. The student similarity group portrait retrieval result is used as context information together with the student information to be predicted as the input of the generation part, thereby generating a description of the student portrait to be predicted and a personalized prediction.

[0145] For obtaining automated conversational communication and support information, including: Generative large language models can automate communication channels, such as chatbots or virtual assistants, to proactively provide students with immediate support and guidance. These assistants based on language large models, such as providing information about academic programs and policies, providing learning tips, and even providing emotional support. Through the above-mentioned student portrait sentiment analysis, etc., the emotions and opinions of students in their growth are analyzed, and the conversation generation capabilities of the generative large language model are used to enable organic human-computer interaction fusion in the portrait system, promoting more personalized and responsive student portrait automatic conversational communication functions.

[0146] To obtain fair evaluation and incentive information, including: Usually, the process of student evaluation is centered on structured data GPA, which often leads to a single evaluation of students. As in algorithm frameworks 1 and 2, the portrait analysis driven by the generative large language model can use the unstructured data in the database system to conduct a comprehensive and fairer evaluation and incentive for student portraits, and promote fairness in the student decision-making process. By incorporating unstructured data such as moral considerations into the development and deployment of the student portrait analysis system, the student portrait evaluation system can be made more fair and diverse.

[0147] Based on the same inventive concept, an embodiment of the present invention also provides a generative human-machine fusion student portrait analysis system.

[0148] Specifically, Figure 4 As shown, the generative human-machine fusion student portrait analysis system provided in this embodiment may include the following modules:

[0149] Multimodal data fusion module, which is used to obtain students' multimodal data, fuse the multimodal data, and extract student portrait features;

[0150] A feature clustering module, which is used to perform clustering processing on student portrait features under the guidance of a generative large language model to obtain multiple groups of student portrait clustering data and their portrait vector databases;

[0151] An information extraction module, which is used to analyze the student portrait clustering data for each group of student portrait clustering data using a generative large language model, and identify the indicator information and risk factor information required for personalized analysis of existing student portraits;

[0152] The result recommendation module is used to take the basic data of the new student to be predicted as input data, and based on the portrait vector data similar to the student to be predicted, use the retrieval-enhanced generative large language model to make personalized portrait predictions for the student to be predicted, and generate corresponding customized recommendation information.

[0153] It should be noted that the steps in the method provided by the present invention can be implemented by using corresponding modules, devices, units, etc. in the system. Those skilled in the art can refer to the technical solution of the method to realize the composition of the system, that is, the embodiments in the method can be understood as preferred examples for constructing the system, which will not be elaborated here.

[0154] An embodiment of the present invention provides a computer terminal, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it can be used to execute any method of the above-mentioned embodiments of the present invention, or to run any system of the above-mentioned embodiments of the present invention.

[0155] Optionally, the memory is used to store programs; the memory may include volatile memory (English: volatile memory), such as random-access memory (English: random-access memory, abbreviated: RAM), such as static random-access memory (English: static random-access memory, abbreviated: SRAM), double data rate synchronous dynamic random access memory (English: Double Data Rate Synchronous Dynamic Random Access Memory, abbreviated: DDR SDRAM), etc.; the memory may also include non-volatile memory (English: non-volatile memory), such as flash memory (English: flash memory). The memory is used to store computer programs (such as applications, functional modules, etc. that implement the above method), computer instructions, etc., and the above computer programs, computer instructions, etc. can be partitioned and stored in one or more memories. And the above computer programs, computer instructions, data, etc. can be called by the processor.

[0156] The above-mentioned computer programs, computer instructions, etc. may be stored in one or more memories in partitions, and the above-mentioned computer programs, computer instructions, data, etc. may be called by a processor.

[0157] The processor is used to execute the computer program stored in the memory to implement the various steps of the method or various modules of the system involved in the above embodiments. For details, please refer to the relevant descriptions in the above method and system embodiments.

[0158] The processor and the memory may be independent structures or integrated structures. When the processor and the memory are independent structures, the memory and the processor may be coupled and connected via a bus.

[0159] An embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, it can be used to execute any method of the above embodiments of the present invention, or to run any system of the above embodiments of the present invention.

[0160] The generative human-machine fusion student portrait analysis method, system, terminal and medium provided by the above embodiment of the present invention utilize the personalized prediction function driven by the large language model, and by analyzing various data such as academic records, extracurricular activities, teacher-student interactions and personal preferences, it is possible to generate customized prediction recommendation data, so that students can make auxiliary references for their academic paths, course selections and employment strategies, thereby improving academic performance and career achievements. The predictive analysis capabilities driven by the large language model are used to support students in early intervention of interference factors. By identifying early indicators of student success or risk factors for academic downturn, proactive intervention measures are taken to provide targeted support and intervention to help students stay on the right track and achieve their academic goals, thereby improving academic results and learning achievements. The resource allocation and operational efficiency functions driven by the large language model help optimize resource utilization by analyzing course selection patterns, student preferences and academic performance data, and make data-driven decisions based on student needs, such as assigning tutors to academic support services or which courses to offer. This resource optimization improves educational operational efficiency and ensures that resources are allocated in a way that maximizes student learning achievements. The use of large language models driven by auxiliary detection and elimination of bias in data and algorithms promotes transparency and fairness in decision-making, while protecting students' privacy and data security, and promoting fair and inclusive practices. By incorporating unstructured data such as moral and political considerations into student evaluation incentives, we ensure that policies are in the best interests of all students and create a more equitable and inclusive learning environment. Driven by large language models, we can achieve personalized learning experiences, early intervention of interference factors, improve academic performance and participation, optimize resource allocation, and fair and inclusive practices. By integrating large language model algorithms, we can achieve enhanced educational outcomes and richer learning experiences.

[0161] All matters not covered in the above embodiments of the present invention are well known in the art.

[0162] The above describes the specific embodiments of the present invention. It should be understood that the present invention is not limited to the above specific embodiments, and those skilled in the art may make various modifications or variations within the scope of the claims, which do not affect the essence of the present invention.

Claims

1. A generative human-computer fusion student portrait analysis method, characterized in that: include: Acquire multimodal data of students, fuse the multimodal data, and extract student portrait features; Providing a generative large language model, performing clustering processing on the student portrait features under the guidance of the generative large language model, and obtaining multiple groups of student portrait clustering data and their portrait vector databases; For each group of the student portrait cluster data, the generative large language model is used to analyze the student portrait cluster data to identify the indicator information and risk factor information required for the personalized analysis of the existing student portraits; The basic data of the new student to be predicted is taken as input data. Based on the portrait vector data similar to the student to be predicted, the retrieval-enhanced generative large language model is used to perform personalized portrait prediction on the student to be predicted and generate corresponding customized recommendation information.

2. The generative human-machine fusion student portrait analysis method according to claim 1 is characterized in that: The step of obtaining multimodal data of students, fusing the multimodal data, and extracting student portrait features includes: Obtain student multimodal data, including: existing structured and unstructured information in the internal school teacher-student system and unstructured information in external resource information; Providing a pre-trained Chinese language model, and using the Chinese language model to vectorize the unstructured information to obtain vector data; Fusing the vector data and the structured information, and performing feature extraction on the fused data to obtain student portrait features; in: The pre-trained Chinese language model is obtained by training an open source language model using structured and unstructured information in an internal school teacher-student system.

3. The generative human-machine fusion student portrait analysis method according to claim 1 is characterized in that: The method provides a generative large language model, performs clustering processing on the student portrait features under the guidance of the generative large language model, and obtains multiple groups of student portrait clustering data and portrait vector databases thereof, including: A generative large language model is provided, and based on the student portrait features, different types of student portrait descriptions are generated using the prompt engineering guidance of the generative large language model; the multimodal data of the student portrait description is encoded into a vector using the generative large language model to obtain a portrait cluster description; The similarity between each student and the portrait cluster description is calculated, and the students are assigned to portrait categories with similarities higher than a set threshold, so as to obtain multiple groups of student portrait cluster data containing structured information and unstructured information, and construct a portrait vector database.

4. The generative human-machine fusion student portrait analysis method according to claim 1 is characterized in that: For each group of the student portrait cluster data, the generative large language model is used to analyze the student portrait cluster data to identify the indicator information and risk factor information required for the personalized analysis of the existing student portraits, including: The understanding and generation capabilities of the generative large language model are used to analyze the structured information and unstructured information in the student portrait clustering data, extract the set features of different portrait clusters, and generate key indicator information and several risk factor information for personalized prediction of existing students and future students.

5. The generative human-machine fusion student portrait analysis method according to claim 1 is characterized in that: The method uses the basic data of the new student to be predicted as input data, and based on the portrait vector data similar to the student to be predicted, uses the retrieval-enhanced generative large language model to perform personalized portrait prediction on the student to be predicted, and generates corresponding customized recommendation information, including: Obtain newly added structured and unstructured information from the internal school teacher-student system as basic data for new students to be predicted; Using the generative large language model and combining it with a retrieval enhancement algorithm, a retrieval-enhanced generative large language model is established; The retrieval part of the retrieval-enhanced generative large language model uses student portrait clustering analysis to retrieve student information related to the student to be predicted in the portrait vector database to obtain a portrait description of a similar student group; The similar student group portrait description is used as context information together with the basic data of the new student to be predicted as the input of the generation part of the retrieval-enhanced generative large language model, and the indicator information and risk factor information of the basic data of the new student to be predicted are predicted, so as to obtain the personalized analysis results of the student portrait to be predicted and generate corresponding customized recommendation information.

6. The generative human-machine fusion student portrait analysis method according to claim 5 is characterized in that: The customized recommendation information includes: personalized portrait prediction recommendation information, evaluation incentive information and portrait key indicator analysis result information.

7. The method for analyzing student portraits using generative human-machine fusion according to any one of claims 1 to 6, characterized in that: Also includes: Use generative large language models to automatically generate dialogues, achieve human-computer interaction fusion, and obtain automated dialogue communication and support information.

8. A generative human-machine fusion student portrait analysis system, characterized in that: include: A multimodal data fusion module, which is used to obtain student multimodal data, fuse the multimodal data, and extract student portrait features; A feature clustering module, which is used to perform clustering processing on the student portrait features under the guidance of a generative large language model to obtain multiple groups of student portrait clustering data and their portrait vector databases; An information extraction module, which is used to analyze the student portrait clustering data using a generative large language model for each group of the student portrait clustering data, and identify the indicator information and risk factor information required for the personalized analysis of the existing student portraits; The result recommendation module is used to take the basic data of the new student to be predicted as input data, and based on the portrait vector data similar to the student to be predicted, use the retrieval-enhanced generative large language model to make personalized portrait predictions for the student to be predicted, and generate corresponding customized recommendation information.

9. A computer terminal comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, it can be used to perform the method described in any one of claims 1 to 7, or to run the system described in claim 8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, it can be used to perform the method described in any one of claims 1 to 7, or to run the system described in claim 8.

Images