College entrance examination learning-up planning system based on Ai intelligent question and answer

Abstract

The invention, which relates to the technical field of artificial intelligence and educational informatization, discloses a college entrance examination learning-up planning system based on Ai intelligent questions and answers, the system comprises a data acquisition module, an intelligent questions and answers interaction module, a knowledge graph construction module, an intelligent recommendation module, and a risk assessment module. The system has the capability of real-time monitoring and dynamic optimization, the risk assessment module can continuously monitor data such as student score fluctuation, college enrollment plan change and professional enrollment popularity change in real time, the uncertainty of application combination is quantified according to the information entropy theory, an ascending planning scheme is dynamically adjusted according to an entropy change threshold value, and the system has the advantages of being high in reliability and high in reliability. Meanwhile, a feedback optimization unit of the intelligent recommendation module can dynamically adjust a recommendation strategy and a fusion weight according to real-time feedback of the students and a system evaluation result, and the continuous optimization and self-adaptive adjustment mechanism can enable an ascending planning scheme to always fit actual conditions of the students and external environment changes.

Description

Technical Field

[0001] The present invention relates to the technical fields of artificial intelligence and educational informatization, and specifically to a college entrance examination admission planning system based on AI intelligent question answering. Background Art

[0002] With the development of social economy and the popularization of higher education, the importance of college entrance examination admission planning has become increasingly prominent. It not only affects whether students can enter their ideal universities and majors, but also has a profound impact on their future career development and personal growth. In the current trend of diversified and personalized education development, a scientific and reasonable admission planning can help students better play to their own advantages and realize personal value. At the same time, college entrance examination admission planning involves a vast amount of information, including the comprehensive strength of universities, professional characteristics, admission rules, employment prospects, etc., as well as students' own interests, hobbies, subject advantages, career inclinations, etc. How to efficiently integrate and utilize this information has become an urgent problem to be solved.

[0003] Currently, there are many deficiencies in the existing college entrance examination admission planning methods on the market. In terms of information integration, traditional admission planning mostly relies on manual collection and collation of information, which is not only inefficient but also difficult to comprehensively cover information such as the constantly changing college enrollment rules, professional dynamics, and employment market demands. In terms of personalized services, existing intelligent assistance systems often simply recommend colleges and majors based on students' grades, lacking in-depth analysis of multi-dimensional factors such as students' interests, hobbies, personality traits, and career inclinations, and unable to truly achieve personalized admission planning. In terms of risk assessment and response, the existing technologies rarely involve it, unable to effectively assess risks brought about by fluctuations in students' grades, changes in college enrollment plans, changes in the popularity of major applications, etc., and also unable to adjust the admission planning scheme in a timely manner according to the risk situation, resulting in great uncertainty for students when filling out college entrance examination applications.

[0004] In summary, the existing college entrance examination admission planning methods are difficult to meet the growing needs of students and parents, and are unable to provide students with a comprehensive, accurate, personalized admission planning service with the ability to respond to risks. With the continuous development and maturity of advanced technologies such as artificial intelligence and big data, applying them to the field of college entrance examination admission planning and developing a system that can integrate a vast amount of information, deeply analyze students' personalized needs, and effectively assess and respond to risks has important practical significance. Summary of the Invention

[0005] The object of the present invention is to make up for the deficiencies of the prior art and provide a college entrance examination admission planning system based on AI intelligent Q&A. It can collect a large amount of information on college major settings, admission cut-off scores over the years, enrollment plans, and employment prospects in real time through a data collection module, and perform efficient data cleaning and preprocessing. At the same time, the knowledge graph construction module integrates this information to construct a dynamically updated domain knowledge graph, which enables the system to provide students with comprehensive, accurate, and timely updated admission information, quickly capture and update relevant information, and students can obtain the latest and most accurate admission information at any time, so as to make more scientific and reasonable decisions. This precise information integration and dynamic update mechanism effectively solves the problems of information lag and inaccuracy in traditional admission planning, providing strong information support for students.

[0006] To solve the above technical problems, the present invention provides the following technical solution: A college entrance examination admission planning system based on AI intelligent Q&A, which consists of: a data collection module, an intelligent Q&A interaction module, a knowledge graph construction module, an intelligent recommendation module, and a risk assessment module; The data collection module collects information on college major settings, admission cut-off scores over the years, enrollment plans, and employment prospects. At the same time, it collects students' personal data in the form of questionnaires, including basic information, academic performance, hobbies, career inclinations, and personality traits; The intelligent Q&A interaction module supports voice and text input through a dialogue engine based on semantic understanding, parses user questions, and generates multi-round dialogues; The knowledge graph construction module integrates structured and unstructured data on admission cut-off scores over the years, college major information, and employment trends to construct a dynamically updated domain knowledge graph; The intelligent recommendation module combines collaborative filtering algorithms with deep learning models to generate a combination of volunteer recommendation lists for colleges and majors based on students' personal data and the knowledge graph; The risk assessment module quantifies the uncertainty of the volunteer combination based on the information entropy theory, dynamically switches between path recombination and locking modes according to the entropy change threshold, and generates a gradient volunteer plan by combining constraint optimization.

[0007] Furthermore, the data collection module includes: a college data collection unit, a student data collection unit, an enrollment data collection unit, and a data preprocessing unit, where: The college data collection unit obtains structured data on college major settings, admission cut-off scores over the years, enrollment plans, and employment prospects in real time and synchronously updates them with the database; The student data collection unit collects students' personal data through dynamic questionnaires and intelligent Q&A interactions, including academic performance, subject hobbies, career inclinations, and personality traits; The enrollment data collection unit updates the enrollment rules in real time, extracts the subject selection requirements, bonus items, and major adjustment notices through semantic parsing; The data preprocessing unit performs denoising, normalization, and correlation mapping on the collected raw data, cleans the missing and abnormal data, and constructs an association index table of students-institutions-enrollment for the knowledge graph module to call.

[0008] Furthermore, the intelligent question-answering interaction module includes: a multimodal input unit, an intention parsing unit, a dynamic retrieval unit, an answer generation and dialogue management unit, where: The multimodal input unit supports multi-source input of voice, text, and images, and converts the student input into a structured query instruction in real time; The intention parsing unit performs semantic understanding based on a semantic understanding model, executes entity recognition, sentiment analysis, and dialogue state tracking, and identifies the type of student questions; The dynamic retrieval unit locates the knowledge nodes and association relationships related to the question in the domain knowledge graph constructed by the knowledge graph construction module according to the semantic understanding result. The retrieval scope includes: institutional and major attribute data, real-time enrollment rule change categories, and historical admission cases of students with the same scores; The answer generation and dialogue management unit generates accurate answers based on the knowledge retrieval results and manages the multi-round dialogue process based on the dialogue strategy.

[0009] Furthermore, the knowledge graph construction module includes: a heterogeneous data integration unit, a graph generation unit, and a dynamic update unit, where: The heterogeneous data integration unit extracts the association index table of students-institutions-enrollment from the data preprocessing unit of the data collection module and constructs three-level attributes; The graph generation unit establishes association relationships based on the three-level attributes to form a dynamic knowledge network. The first layer is the subordinate relationship between institutional nodes and major nodes, that is, the institutions to which each major belongs. The second layer is the numerical constraint relationship between major nodes and admission cut-off scores and enrollment plans, that is, the admission cut-off scores and enrollment plan numbers of a major play a constraining role in the application and admission of that major. The third layer is the dynamic influence relationship between enrollment rule nodes and major adjustments and subject selection requirements, that is, changes in the rules lead to the opening, adjustment of majors, and changes in subject selection requirements; The dynamic update unit updates the graph node attributes by monitoring the enrollment rule changes, score line fluctuations, and employment trend data captured by the data collection module in real time, and reconstructs the dynamic graph path nodes through the threshold of the entropy flow diagnosis. The threshold, where , where is the entropy change value of the knowledge graph, is the entropy value of the volunteer combination of the risk assessment module, is the change amount of the enrollment rule score line, represents the number of variables participating in the calculation. When the graph entropy change threshold is reached, the graph version management mechanism is started, and the time series features are embedded and the dynamic graph path nodes are reconstructed.

[0010] Furthermore, the three-level attributes of the heterogeneous data integration unit in the knowledge graph construction module specifically include: Entity layer: including entity types such as institutions, majors, occupations, and subject ability labels; Relationship layer: including semantic relationships such as major - required subject correlation degree, institution - regional employment index; Dynamic weight layer: adjusts the reference weight of historical data based on the time decay factor, that is , is the current weight, is the initial weight, is the decay coefficient, and are the current time and the historical data time respectively.

[0011] Furthermore, the intelligent recommendation module includes: a feature vector generation unit, a hybrid recommendation unit, a candidate set generation unit, and a feedback optimization unit, where: The feature vector generation unit extracts the institution feature vector , student portrait vector and enrollment rule constraint vector based on the dynamic knowledge network output by the knowledge graph construction module, and constructs a three-dimensional feature space; The hybrid recommendation unit adopts a parallel computing architecture. The first channel calculates the similarity between the student and historical admission cases based on collaborative filtering , the second channel learns the implicit association of institution - major - student , the third channel uses the knowledge graph to mine the reachable paths under the enrollment rule constraints , and the results of the three channels are weighted and fused to calculate the confidence to generate an initial recommendation pool, where is the dynamic fusion weight and ; The candidate set generation unit performs multi-dimensional pruning on the initial recommendation pool and outputs a candidate set of institution - major combinations sorted by confidence ; The feedback optimization unit dynamically adjusts the fusion weight according to the student's real-time feedback and the system evaluation results, and optimizes the recommendation results.

[0012] Furthermore, the risk assessment module includes: an entropy value calculation unit, a threshold judgment unit, a path recombination unit, a constraint unit, and a multi-objective optimization unit, where: The entropy calculation unit is based on the dynamic weight of the knowledge graph construction module And the candidate set of the recommendation module, calculate the volunteer combination entropy value ,in, represents the i-th volunteer The probability of admission, is the weight of the i-th volunteer in the knowledge graph, is the confidence score of the i-th volunteer in the recommendation module, and n represents the total number of candidate colleges and majors in the current volunteer combination; The threshold judgment unit compares the student's fitness score AS with the preset risk threshold. When When divided into retainable volunteers, the , where γ is the constraint capacity coefficient of the enrollment rules; The path reorganization unit and Satisfy it for a while, To maximize the entropy value of the volunteer combination, based on the association between colleges and majors in the knowledge graph, we retrieve replacement nodes with the same gradient from the knowledge graph, generate a new path set, and remove nodes that do not meet the dynamic constraints. The constraint unit constructs constraint conditions based on the new path set generated by the path reorganization unit and calculates the entropy value of the new path , filter to meet the constraints: Volunteer Program, The minimum entropy reduction for path reorganization; The multi-objective optimization unit constructs a multi-objective optimization function, simultaneously optimizes the admission probability and the risk entropy value, and optimizes the volunteer plan under constraints.

[0013] Furthermore, the multi-objective optimization function of the multi-objective optimization unit is: , Indicates that it is limited to, this function indicates that it is limited to Under the condition of , the total admission probability of the optimized volunteer combination is maximized, where The number of volunteers retained after optimization, represents the i-th volunteer probability of admission.

[0014] Compared with the existing technology, this college entrance examination planning system based on AI intelligent question answering has the following beneficial effects: 1. The college entrance examination planning system of the present invention has the ability of real-time monitoring and dynamic optimization. The risk assessment module will continuously monitor data such as student performance fluctuations, changes in college enrollment plans, and changes in the popularity of professional applications in real time. It will quantify the uncertainty of the volunteer combination based on the information entropy theory and dynamically adjust the enrollment planning scheme based on the entropy change threshold. At the same time, the feedback optimization unit of the intelligent recommendation module will dynamically adjust the recommendation strategy and fusion weight based on the real-time feedback of students and the results of system evaluation. This continuous optimization and adaptive adjustment mechanism can make the enrollment planning scheme always fit the actual situation of the students and changes in the external environment, ensure that the students' enrollment planning is always in the optimal state, and maximize the protection of students' interests.

[0015] 2. The college entrance examination planning system of the present invention can collect massive information such as university major settings, admission scores in previous years, enrollment plans, employment prospects, etc. in real time through the data acquisition module, and perform efficient data cleaning and preprocessing. At the same time, the knowledge graph construction module integrates this information and constructs it into a dynamically updated domain knowledge graph, which enables the system to provide students with comprehensive, accurate and timely updated enrollment information.

[0016] Other advantages, objects and features of the present invention will be described in part in the following description and, in part, will be apparent to those skilled in the art based on an examination of the following or may be learned from the practice of the invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] To more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. Those skilled in the art can also derive other drawings based on these drawings without inventive effort.

[0018] Figure 1 This is an operational flow chart of a college entrance examination planning system based on AI intelligent question answering; Figure 2 This is a schematic diagram of the module composition of a college entrance examination planning system based on AI intelligent question answering. DETAILED DESCRIPTION

[0019] In order to further illustrate the technical means and effects adopted by the present invention to achieve the predetermined purpose of the invention, the specific implementation methods, structures, features and effects of the present invention are described in detail below in conjunction with the accompanying drawings and preferred embodiments.

[0020] Example 1: Figure 2As shown, this embodiment focuses on a college entrance examination planning system based on AI intelligent question and answer, and elaborates on its practical application in the college entrance examination planning scenario. The system integrates modules such as data collection, intelligent question and answer interaction, knowledge graph construction, intelligent recommendation and risk assessment, deeply integrates college and student data, and realizes personalized and intelligent enrollment planning services. By showing that the various modules of the system work together, it provides students with accurate college and major recommendations, intelligent question and answer, and risk management.

[0021] The data collection module collects information on college majors, admission scores over the years, enrollment plans, and employment prospects. At the same time, it uses questionnaires to collect students' personal data, including basic information, academic performance, interests and hobbies, career tendencies, and personality traits. The module includes: a college data collection unit, a student data collection unit, an enrollment data collection unit, and a data preprocessing unit. Among them, the college data collection unit automatically collects educational data at time intervals. During the visit, it accurately identifies and extracts structured data such as college majors, admission scores over the years, enrollment plans, and employment prospects. The collected data is updated synchronously to ensure the timeliness and accuracy of the data; the student data collection unit uses multiple methods to collect students' personal data. On the one hand, by displaying a dynamic questionnaire at the front end of the system, the questionnaire content will be intelligently adjusted according to the student's previous answers. For example, if the student initially answers that he is interested in science and engineering, the subsequent questionnaire will specifically ask more detailed science and engineering related interests. On the other hand, with the help of the intelligent question-answering interaction module, information is collected during the dialogue between the student and the system. For example, when a student asks questions about computer science, The system will also record the content and method of students' questions, and analyze their academic performance, subject interests, career inclinations, personality traits and other information. By combining these two methods, the system can obtain students' personal data as comprehensively as possible, providing a basis for subsequent personalized analysis and recommendations; the admissions data collection unit is responsible for updating the admissions rules in real time. When new admissions rules are released, the system uses semantic parsing technology to conduct in-depth analysis of the rule text. For example, it extracts subject selection requirements, determines the specific requirements of different majors for students' selected subjects, identifies bonus items, clarifies under what circumstances students can get bonus points, interprets major adjustment notices, and understands the addition, cancellation or adjustment of university majors. This information is crucial for accurately planning the path to further study. The admissions data collection unit ensures that students and parents can obtain the latest admissions rules information in a timely manner; the data preprocessing unit processes the collected raw data, eliminates non-compliant data, and performs normalization processing to unify data of different magnitudes to the same scale for subsequent calculation and analysis. By associating students' personal data, school information and admissions rules, it provides a structured data foundation for the knowledge graph construction module.

[0022] The intelligent Q&A interaction module supports voice and text input through a dialogue engine based on semantic understanding, parses users' questions and generates multi-turn conversations. This module includes: a multi-modal input unit, an intention parsing unit, a dynamic retrieval unit, and an answer generation and dialogue management unit. Among them: When the student interacts with the system, the multi-modal input unit plays a role. This unit supports multi-source input of voice, text, and images. When the student uses voice input, the system uses speech recognition technology to convert the voice signal into text information. If the student directly inputs text, the text content is directly received. For image input, the system extracts the text information or relevant features therein through image recognition technology, and converts these input contents into structured query instructions to provide a standardized data format for subsequent intention parsing; The intention parsing unit deeply analyzes the structured query instructions based on an advanced semantic understanding model. In this process, entity recognition is performed to determine the key entities in the instructions, such as college names, major names, scores, etc., sentiment analysis is carried out to judge the student's attention level and sentiment tendency towards the question, and dialogue state tracking is carried out to record the student's previous questions and the system's answers in order to understand the context of the entire conversation and accurately identify the type of the student's question, such as whether it is about college recommendation, major introduction, or admission rule consultation, etc. The accuracy of intention parsing directly affects the quality of the subsequent answer generation. If the intention parsing is incorrect, it may lead to the provided answer not meeting the student's needs; The dynamic retrieval unit retrieves in the domain knowledge graph constructed by the knowledge graph construction module according to the results of the intention parsing unit. The retrieval scope includes college and major attribute data, such as the comprehensive rankings of colleges and the course settings of majors, real-time rule change categories, such as the latest admission rule adjustments; Historical admission cases in the same score range, so as to provide more reference information for students. In this way, the knowledge nodes and association relationships related to the question are quickly located to provide rich data support for answer generation; The answer generation and dialogue management unit generates accurate answers according to the knowledge retrieval results obtained by the dynamic retrieval unit. If the retrieved information is relatively simple, such as the basic introduction of a major, the information is directly sorted into a clear and easy-to-understand text output; If it involves complex problems, such as the comparative analysis of multiple colleges, a detailed analysis report is generated through logical reasoning and information integration. At the same time, based on the dialogue strategy, the multi-turn dialogue process is managed. The system can continue to provide relevant information according to the previous dialogue records and the current question to maintain the coherence and effectiveness of the dialogue.

[0023] The knowledge graph construction module integrates structured and unstructured data of admission cut-off scores, institution major information, and employment trends over the years to construct a dynamically updated domain knowledge graph. This module includes: a heterogeneous data integration unit, a graph generation unit, and a dynamic update unit. Among them, the heterogeneous data integration unit extracts the student-institution-admission correlation index table from the data preprocessing unit of the data collection module, constructs three-level attributes. At the entity layer, entity types including institutions, majors, occupations, subject ability labels, etc. are determined. These entities are the basic elements of the knowledge graph and provide a basis for subsequent relationship construction. For example, each university is regarded as an independent entity, including attributes such as its name, address, and school-running level. The relationship layer includes semantic relationships such as major-required subject correlation degree, institution-regional employment index, etc. The major-required subject correlation degree is determined by analyzing the major curriculum settings and subject requirements, and is represented by a numerical value. The higher the value, the greater the dependence of the major on the subject. The institution-regional employment index is calculated by statistically analyzing the employment situation of the school's graduates in a specific region, reflecting the employment competitiveness of the institution in that region; the dynamic weight layer adjusts the reference weight of historical data based on the time decay factor, and the formula is: . Among them, is the current weight, is the initial weight, is the decay coefficient, which is preset by the system according to the timeliness and importance of the data, and are the current time and the historical data time respectively. In this way, the latest data has a higher weight, ensuring that the knowledge graph can reflect the latest situation; the graph generation unit establishes association relationships based on the three-level attributes. The graph generation unit forms a dynamic knowledge network. The first layer is the subordinate relationship between institution nodes and major nodes, clarifying which institution each major belongs to. The second layer is the data constraint relationship between major nodes and admission cut-off scores and enrollment plans. The admission cut-off scores and enrollment plan numbers play a key constraint role in the application and admission of majors. When the major admission cut-off score is high and the enrollment plan is small, the difficulty of applying for this major is relatively large. The third layer is the dynamic influence relationship between rule nodes and major adjustment and subject selection requirements. Through these three layers of relationships, a comprehensive and dynamic knowledge network is constructed, providing rich knowledge support for the system; the dynamic update unit captures rule changes, score line fluctuations, and employment trend data by real-time monitoring the data collection module. When data changes are detected, the entropy change value of the knowledge graph is calculated. Among them, is the entropy change value of the knowledge graph, is the entropy value of the volunteer combination of the risk assessment module, is the change amount of the admission rule cut-off score, represents the number of variables participating in the calculation. When When the graph entropy change threshold is reached, the graph version management mechanism is activated, temporal features are embedded, and the dynamic graph path nodes are reconstructed to ensure that the knowledge graph can accurately reflect the latest situation, providing a reliable basis for subsequent intelligent recommendation and risk assessment.

[0024] The intelligent recommendation module combines collaborative filtering algorithms with deep learning models to generate a combined list of college and major volunteer recommendations based on students' personal data and the knowledge graph. This module includes: a feature vector generation unit, a hybrid recommendation unit, a candidate set generation unit, and a feedback optimization unit. Among them: The feature vector generation unit constructs a three-dimensional feature space by extracting college feature vectors , student portrait vectors and enrollment rule constraint vectors . When extracting college feature vectors, information such as the comprehensive ranking of colleges, the employment rate of graduates, and the application popularity in the region where they are located is obtained from the knowledge graph; for student portrait vectors, data such as grades, hobbies, and career inclinations collected by the student data collection unit are combined; the constraint vectors extract key information such as subject selection requirements and enrollment restrictions from the enrollment rules obtained by the enrollment data collection unit. By constructing this three-dimensional feature space, the key information of colleges, students, and rules is integrated together, providing comprehensive data support for subsequent recommendation algorithms; The hybrid recommendation unit adopts a parallel computing architecture and works collaboratively through three channels. The first channel calculates the similarity between students and historical admission cases based on collaborative filtering . By calculating the similarity between the student portrait vector and the student portrait vectors in historical admission cases, historical cases similar to the current student are found, and the admission situations of these cases are referred to for recommendations. The second channel learns the implicit associations between colleges - majors - students through deep learning models . The deep neural network is used to mine the potential relationships in the data to predict students' preferences for different college majors. The third channel uses the knowledge graph to mine the reachable paths under rule constraints . Eligible college and major paths are screened according to requirements. Finally, the results of the three channels are weighted and fused to calculate the confidence to generate an initial recommendation pool, where is the dynamic fusion weight and , these weights will be dynamically adjusted according to the system's evaluation and the students' feedback to ensure the accuracy and rationality of the recommendation results; the candidate set generation unit performs multi-dimensional pruning on the initial recommendation pool. First, it executes gradient matching. Based on the students' grades and the admission cut-off scores of the institutions, it reasonably arranges institutions at different gradients, such as sprint institutions, safe institutions, and guarantee institutions, to ensure that the recommended institutions are both challenging and have a certain admission probability. Secondly, it conducts cold start compensation. For newly enrolled students or those with less data, it supplements the recommendation results through the recommendation results of similar students to avoid overly single or inaccurate recommendation results. Finally, it conducts risk hedging. Combining the results of the risk assessment block, it adjusts high-risk institutions and majors to reduce the overall risk. After these operations, it outputs a candidate set of institution-major combinations sorted by confidence to provide students with diverse choices; the feedback optimization unit dynamically adjusts the fusion weights based on the students' real-time feedback and the system's evaluation results. When the students are satisfied with the recommendation results, it appropriately increases the weight of the current recommendation strategy. If the students are not satisfied with the recommendation results, it analyzes the students' feedback content, finds the gap between the recommendation results and the students' needs, and adjusts the fusion weights accordingly.

[0025] The risk assessment module quantifies the uncertainty of the volunteer combination based on the information entropy theory, dynamically switches the path recombination and locking modes according to the entropy change threshold, and generates a gradient volunteer plan by combining constraint optimization. This module includes: an entropy value calculation unit, a threshold judgment unit, a path recombination unit, a constraint unit, and a multi-objective optimization unit. Among them: the entropy value calculation unit calculates the entropy value of the volunteer combination based on the dynamic weights of the knowledge graph construction module and the candidate set of the recommendation module , where represents the admission probability of the i-th volunteer , is the weight of the i-th volunteer in the knowledge graph, is the confidence score of the i-th volunteer in the recommendation module, and n represents the total number of candidate institutions and majors in the current volunteer combination. Through this formula, considering the weights in the knowledge graph and the confidence of the recommendation module, it quantifies the uncertainty of the volunteer combination and provides data support for subsequent risk judgment; the threshold judgment unit conducts risk judgment based on the comparison result between the student's fitness score A5 and the preset risk threshold; the fitness score formula is , where γ is the rule constraint ability coefficient, which is set by the system according to the importance and influence degree. When , it is classified as a high-risk volunteer; when , it is classified as a volunteer that can be retained. In this way, it quickly judges the risk level of the volunteer and provides a basis for subsequent path recombination and optimization; the path recombination unit and when either is satisfied is the maximum value of the entropy of the volunteer combination. The path recombination unit retrieves alternative nodes of the same gradient from the knowledge graph based on the association relationship between institutions and majors in the knowledge graph, generates a new path set, eliminates nodes that do not meet the dynamic constraint conditions, and reconstructs the volunteer combination path to reduce the overall risk; the constraint unit constructs constraint conditions according to the new path set generated by the path recombination unit and calculates the entropy value of the new path , and filters out the volunteer plan that meets the constraint condition: where is the minimum entropy reduction amount for path recombination. Through this constraint condition, it is ensured that the volunteer plan after path recombination has an actual effect in reducing risk and avoids ineffective path recombination; the multi-objective optimization unit constructs a multi-objective optimization function: , where represents maximizing the total admission probability of the optimized volunteer combination under the condition of being restricted by , is the number of volunteers retained after optimization, represents the admission probability of the i-th volunteer . Through this optimization function, on the premise of ensuring controllable risk, the probability of students being admitted to ideal institutions and majors is increased as much as possible, realizing the balanced optimization of multiple objectives.

[0026] In summary, this embodiment comprehensively and deeply demonstrates the operating mechanism of the college entrance examination admission planning system based on AI intelligent Q&A. Starting from the data collection module, the institution, student, and enrollment data collection units each perform their own functions, collecting data on various aspects such as college major settings, student personal information, and enrollment rules respectively. The data preprocessing unit further cleans, normalizes, and correlates this data, laying a solid data foundation for the entire system and ensuring the accuracy of subsequent analysis and decision-making. The intelligent Q&A interaction module receives diverse input methods from students through the multi-modal input unit. The intent parsing unit accurately understands the needs of students, the dynamic retrieval unit obtains relevant information from the knowledge graph, and the answer generation and dialogue management unit provides accurate and coherent answers accordingly, achieving efficient and intelligent human-computer interaction and meeting the information query needs of students during the admission planning process. The heterogeneous data integration unit of the knowledge graph construction module constructs a three-level attribute including an entity layer, a relationship layer, and a dynamic weight layer. The graph generation unit builds a dynamic knowledge network based on this, and the dynamic update unit monitors data changes in real time, determining whether to update the graph by calculating the entropy change value to ensure the timeliness and accuracy of the knowledge graph, providing rich and continuously updated knowledge support for the system. The feature vector generation unit of the intelligent recommendation module extracts key features to construct a three-dimensional feature space. The hybrid recommendation unit uses parallel computing and multi-channel fusion to generate an initial recommendation pool. The candidate set generation unit performs multi-dimensional pruning, and the feedback optimization unit adjusts the recommendation strategy based on feedback, providing personalized and precise college and major recommendations for students and improving the pertinence and effectiveness of the recommendations. The entropy value calculation unit of the risk assessment module quantifies the uncertainty of the volunteer combination, the threshold judgment unit divides the risk levels, the path recombination unit adjusts high-risk volunteers, the constraint unit filters effective volunteer plans, and the multi-objective optimization unit maximizes the admission probability under the premise of controllable risk, achieving effective management of volunteer risks and optimization of the admission planning plan. Each module of the entire system closely collaborates and cooperates with each other, forming a complete ecological closed-loop, which can help students comprehensively understand information related to the college entrance examination admission, and make scientific and reasonable volunteer decisions based on their own situations.

[0027] Embodiment 2: As Figure 1 shown, this embodiment provides an operation process for students to use a college entrance examination admission planning system based on AI intelligent Q&A for college entrance examination volunteer filling and admission planning. The specific steps of this operation process are as follows: Improve personal information: After entering the system, on the personal information page, the student fills in basic information as required, such as name, gender, region, etc. Then, the student details the academic performance, covering usual grades and mock exam grades of each subject. At the same time, through questionnaires or Q&A forms provided by the system, the student carefully fills in their hobbies, career inclinations, and personality traits.

[0028] Query information: Students can use the intelligent Q&A interaction module to input questions in the form of voice or text. After the system analyzes the questions, it retrieves relevant information from the knowledge graph and gives accurate answers to help students understand various information about colleges and majors.

[0029] Obtain preliminary recommendations: The system generates a preliminary list of college and major volunteer recommendations based on the personal information and grades entered by the students, combined with the knowledge graph and the intelligent recommendation module.

[0030] Risk assessment and plan adjustment: The risk assessment module generates candidate volunteer combinations based on the dynamic weights of the knowledge graph and the recommendation module. It conducts a risk assessment on the volunteer combinations. The entropy calculation unit calculates the entropy value of the volunteer combinations. The threshold judgment unit determines the volunteer risk level based on the comparison result between the student's fitness score and the preset risk threshold. If there are high-risk volunteers, the path reorganization unit retrieves alternative nodes of the same gradient based on the association relationship between colleges and majors in the knowledge graph, generates a new set of paths, and eliminates nodes that do not meet the dynamic constraint conditions. The constraint unit constructs constraint conditions based on the new set of paths and filters out volunteer plans that meet the conditions. The multi-objective optimization unit optimizes the volunteer plan on the premise of ensuring controllable risks, and regenerates a more reasonable new volunteer combination to recommend to the students.

[0031] Deeply understand the details of the volunteers: For the recommended colleges and majors, students click to view the detailed information, including the comprehensive strength of the colleges, the professional curriculum settings, the employment prospects, the trend of the admission cut-off scores over the years, etc., to further understand the feasibility and suitability of the volunteers.

[0032] Determine the volunteer filling plan: After multiple rounds of viewing and adjustment by the students, the final volunteer filling plan is determined. Before confirmation, the system conducts a risk reminder and a volunteer rationality check again to ensure that the plan meets the needs of the students and the local college entrance examination volunteer filling rules.

[0033] Save and export the plan: After confirmation, the students save the volunteer filling plan. At the same time, the system supports exporting the plan.

[0034] The above is only a preferred embodiment of the present invention, and it is not intended to limit the present invention in any form. Although the present invention has been disclosed above with the preferred embodiment, it is not intended to limit the present invention. Any person skilled in the art can make some changes or modifications to the above-disclosed technical content to make equivalent embodiments of equivalent changes. However, as long as it does not depart from the technical content of the present invention, any brief modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention still fall within the scope of the technical solution of the present invention.

Claims

1. An entrance examination planning system for college entrance based on AI intelligent Q&A, characterized in that, The system consists of: a data collection module, an intelligent Q&A interaction module, a knowledge graph construction module, an intelligent recommendation module, and a risk assessment module; The data collection module collects information on college major settings, admission scores over the years, enrollment plans, and employment prospects. At the same time, it collects students' personal data in the form of questionnaires, including basic information, academic performance, hobbies, career inclinations, and personality traits; The intelligent Q&A interaction module supports voice and text input through a dialogue engine based on semantic understanding, parses users' questions, and generates multi-round conversations; The knowledge graph construction module integrates structured and unstructured data on admission scores over the years, college major information, and employment trends to construct a dynamically updated domain knowledge graph; The intelligent recommendation module combines collaborative filtering algorithms with deep learning models to generate a combination of volunteer recommendation lists for colleges and majors based on students' personal data and the knowledge graph; The risk assessment module quantifies the uncertainty of the volunteer combination based on the information entropy theory, dynamically switches between path recombination and locking modes according to the entropy change threshold, and generates a gradient volunteer plan by combining constraint optimization.

2. The college entrance examination admission planning system based on AI intelligent Q&A according to claim 1, characterized in that, The data collection module includes: a college data collection unit, a student data collection unit, an enrollment data collection unit, and a data preprocessing unit, where: The college data collection unit obtains structured data on college major settings, admission scores over the years, enrollment plans, and employment prospects in real time and synchronously updates them with the database; The student data collection unit collects students' personal data, including academic performance, subject hobbies, career inclinations, and personality traits, through dynamic questionnaires and intelligent Q&A interactions; The enrollment data collection unit updates the enrollment rules in real time and extracts subject selection requirements, bonus items, and major adjustment notices through semantic parsing; The data preprocessing unit denoises, normalizes, and performs correlation mapping on the collected raw data, cleans missing and abnormal data, and constructs an association index table of students - colleges - enrollment for the knowledge graph module to call.

3. The college entrance examination entrance planning system based on AI intelligent Q&A according to claim 1, characterized in that, The intelligent Q&A interaction module includes: a multimodal input unit, an intent parsing unit, a dynamic retrieval unit, and an answer generation and dialogue management unit, where: The multimodal input unit supports multi-source input of voice, text, and images, and real-time converts students' input into structured query instructions; The intent parsing unit performs semantic understanding based on a semantic understanding model, executes entity recognition, sentiment analysis, and dialogue state tracking, and identifies the types of students' questions; The dynamic retrieval unit locates knowledge nodes and association relationships related to the question in the domain knowledge graph constructed by the knowledge graph construction module according to the semantic understanding result. The retrieval scope includes: college and major attribute data, real-time enrollment rule change categories, and historical admission cases in the same score range; The answer generation and dialogue management unit generates accurate answers based on the knowledge retrieval result and manages the multi-round dialogue process based on dialogue strategies.

4. The college entrance examination entrance planning system based on AI intelligent Q&A according to claim 1, wherein The knowledge graph construction module includes: a heterogeneous data integration unit, a graph generation unit, and a dynamic update unit, where: The heterogeneous data integration unit extracts the association index table of students - colleges - enrollment from the data preprocessing unit of the data collection module and constructs three-level attributes; The spectrum generation unit establishes an association relationship based on three-level attributes to form a dynamic knowledge network. The first level is the subordination relationship between the institution node and the major node, that is, the institution to which each major belongs. The second level is the numerical constraint relationship between the major node, the admission score line, and the enrollment plan, that is, the admission score line and the enrollment plan number of the major play a constraining role in the application and admission of the major. The third level is the dynamic influence relationship between the enrollment rule node and the major adjustment and subject selection requirements, that is, the change of the rule leads to the opening, adjustment of the major and the change of subject selection requirements; The dynamic update unit captures data on changes in enrollment rules, fluctuations in cut-off scores, and employment trends through real-time monitoring of the data acquisition module, updates the attributes of the graph nodes, and reconstructs the path nodes of the dynamic graph by combining the threshold of entropy flow diagnosis, and the , where is the entropy change value of the knowledge graph, is the entropy value of the volunteer combination of the risk assessment module, is the change amount of the cut-off score of the enrollment rule, represents the number of variables participating in the calculation. When the entropy change threshold of the graph is reached, the graph version management mechanism is started, and the time series features are embedded and the path nodes of the dynamic graph are reconstructed.

5. The college entrance examination entrance planning system based on AI intelligent Q&A according to claim 4, characterized in that, The three-level attributes of the heterogeneous data integration unit in the knowledge graph construction module specifically include: Entity layer: including entity types such as institutions, majors, occupations, and subject ability labels; Relationship layer: including semantic relationships such as major-required subject correlation degree and institution-regional employment index; Dynamic weight layer: Adjust the reference weight of historical data based on the time decay factor, that is , is the current weight, is the initial weight, is the decay coefficient, and are the current time and the historical data time respectively.

6. The college entrance examination entrance planning system based on AI intelligent Q&A according to claim 1, characterized in that The intelligent recommendation module includes: a feature vector generation unit, a hybrid recommendation unit, a candidate set generation unit, and a feedback optimization unit, where: The feature vector generation unit extracts the institutional feature vectors based on the dynamic knowledge network output by the knowledge graph construction module , the student portrait vectors and the enrollment rule constraint vectors to construct a three-dimensional feature space; The hybrid recommendation unit adopts a parallel computing architecture. The first channel calculates the similarity between students and historical admission cases based on collaborative filtering , the second channel learns the implicit association of colleges - majors - students , the third channel uses the knowledge graph to mine the reachable paths under the constraints of enrollment rules , and the results of the three channels are weighted and fused to calculate the confidence to generate an initial recommendation pool, where is the dynamic fusion weight and ; The candidate set generation unit performs multi-dimensional pruning on the initial recommendation pool and outputs a candidate set of college-major combinations sorted by confidence ; The feedback optimization unit dynamically adjusts the fusion weight according to the real-time feedback of students and the system evaluation results to optimize the recommendation results.

7. The college entrance examination entrance planning system based on AI intelligent Q&A according to claim 1, characterized in that, The risk assessment module includes: an entropy value calculation unit, a threshold judgment unit, a path recombination unit, a constraint unit, and a multi-objective optimization unit, where: The entropy value calculation unit calculates the entropy value of the volunteer combination based on the dynamic weights of the knowledge graph construction module and the candidate set of the recommendation module , where represents the admission probability of the i-th volunteer , is the weight of the i-th volunteer in the knowledge graph is the confidence score of the i-th volunteer in the recommendation module, and n represents the total number of candidate colleges and majors in the current volunteer combination; The threshold judgment unit classifies it as a high-risk volunteer according to the comparison result between the adaptability score AS of the student and the preset risk threshold. When it is classified as a high-risk volunteer, and when it is classified as a volunteer that can be retained. The , where γ is the enrollment rule constraint ability coefficient. When the path recombination unit and meet one of them, is the maximum value of the volunteer combination entropy. Based on the association relationship between institutions and majors in the knowledge graph, retrieve alternative nodes of the same gradient from the knowledge graph, generate a new path set, and eliminate nodes that do not meet the dynamic constraint conditions; The constraint unit constructs constraint conditions based on the new path set generated by the path recombination unit and calculates the entropy value of the new path , and filters the volunteer plans that meet the constraint conditions: where is the minimum entropy value reduction for path recombination; is the minimum entropy value reduction for path recombination; The multi-objective optimization unit constructs a multi-objective optimization function to synchronously optimize the admission probability and the risk entropy value, and optimizes the volunteer plan under the constraint conditions.

8. The college entrance examination entrance planning system based on AI intelligent Q&A according to claim 7, characterized in that, The multi-objective optimization function of the multi-objective optimization unit is as follows: , denoted as subject to, this function means to maximize the total admission probability of the optimized volunteer combination subject to , where is the number of volunteers retained after optimization, represents the i-th volunteer and its admission probability.

Images