College entrance examination application filling auxiliary system

Abstract

The invention belongs to the technical field of application filling, and particularly relates to a college entrance examination application filling auxiliary system which comprises a user input module, a plug-in calling module, a knowledge base system, a data construction and analysis module, a semantic interaction configuration module and an answer synthesis and feedback module. According to the system, a college entrance examination voluntary reporting AI Agent is constructed on the basis of a Baidu cloud sail platform App Builder, a self-defined knowledge base and multiple college entrance examination plug-in tools are fused, and a voluntary auxiliary platform with knowledge driving, semantic question answering and intelligent recommendation is constructed in combination with large model reasoning ability; the platform is devoted to structurally integrate complicated policy requirements and college information, the understanding cost of examinees and parents is reduced, the scientificity and efficiency of volunteer filling are improved, meanwhile, the platform has good expansibility, and the platform can adapt to other new college entrance examination provinces throughout the country.

Description

Technical Field

[0001] This invention relates to the field of college application technology, specifically a college application assistance system. Background Technology

[0002] With the comprehensive advancement of China's college entrance examination system reform, especially after the implementation of the "new college entrance examination 3+1+2" model in provinces such as Inner Mongolia, the information structure faced by candidates in the process of filling out their college applications has become increasingly complex.

[0003] Traditional college application methods rely heavily on the experience and guidance of parents or teachers, which has several drawbacks: 1. Subject combinations have a wide impact on the range of majors: Different universities and majors have strict requirements for subject selection; if a student's combination doesn't match, they will be disqualified from applying. 2. Score and ranking assessments are not intuitive: Students often only know their college entrance examination scores but are unclear about their corresponding rankings and their relationship with previous years' university admission lines. 3. Information channels are fragmented: Students need to consult multiple channels, such as the education examination authority, university websites, and third-party platforms, increasing the burden of information filtering. 4. Lack of personalized decision support: Most current recommendation systems are based solely on score matching, ignoring students' subject preferences, regional preferences, and future plans. Furthermore, while previous college applications were submitted in batches using parallel choices, the current new college application system does not use batches; students submit 45 parallel choices at once, making it more difficult and increasing the risk of being rejected. This necessitates the assistance of big data and artificial intelligence.

[0004] Therefore, a college entrance examination application assistance system is proposed to solve the problems mentioned above. Summary of the Invention

[0005] The purpose of this invention is to provide a college entrance examination application assistance system to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a college entrance examination application assistance system, comprising a user input module, a plug-in calling module, a knowledge base system, a data construction and analysis module, a semantic interaction configuration module, and an answer synthesis and feedback module;

[0007] The user input module is the entry point for all interactions. Its core task is to receive the user's initial input, including but not limited to college entrance examination scores, subject combinations, regional and university preferences, and keywords related to major intentions. It supports a combination of free text and structured input. The system uses Prompt engineering and contextual understanding to parse natural language into parameterized information and dynamically fill it into the semantic request template of the model.

[0008] The plugin calling module is responsible for expanding the knowledge boundary and data retrieval capabilities of the model. It integrates the native plugin capabilities provided by Baidu Cloud Qianfan and currently calls three types of plugins: Baidu College Entrance Examination Plugin, Baidu AI Search Plugin, and Code Translator Plugin.

[0009] The knowledge base system undertakes the basic support work for the system's interpretability, reasonability, and high-precision recommendation tasks. The system consists of two core sub-bases.

[0010] The data construction and analysis module is the starting point for building the system's knowledge source. It is responsible for the programmatic collection, structured cleaning and model reorganization of external public data resources, as well as the supply of high-quality training data.

[0011] The semantic interaction configuration module is located upstream in the large language model call chain and is responsible for imposing instruction constraints and style training on the model.

[0012] The response synthesis and feedback module is the intersection of all model inputs and knowledge retrieval. It receives the main inference results of the model, the feedback data from the plugins, and the matching results from the knowledge base. It then refines the language, standardizes the format, and organizes the logic to ultimately form a structured and understandable output.

[0013] Preferably, the user input module also integrates a user role recognition function. The system determines whether the user is a student, parent, or teacher / advisor based on their language style, and then automatically adjusts the dialogue style and recommendation method. When a user enters the system for the first time, they will receive an initial opening speech to guide them to input key parameters, reducing the operational threshold and enhancing the perception of personalized services. The user input module is not only the entry point for input, but also the starting point for maintaining the context. The model maintains the user's input state through multi-turn semantic association, realizing state maintenance and task tracking in long dialogue scenarios.

[0014] Preferably, in the plugin calling module, the Baidu College Entrance Examination plugin is used to query authoritative information on universities, majors, historical admission data, and educational levels; the Baidu AI Search plugin is used for high-level semantic question answering tasks, such as retrieving unstructured document content such as admission brochures, professional training programs, employment trends, and national policies, and the model performs language summarization and answers; the code translator plugin is used to execute custom data processing scripts, such as web crawling, data cleaning, field alignment, and data structure transformation, enabling the Agent to have the ability to evolve and adjust the knowledge base structure and content as needed during runtime.

[0015] The plugin module is tightly integrated with the model invocation logic and has the ability to make multi-round collaborative calls. That is, when the model fails to obtain the answer from the knowledge base when the user asks a question, the model will automatically choose whether to call the plugin and integrate the returned results.

[0016] Preferably, in the knowledge base system, the two core sub-bases include a score-segment database and a subject selection requirements database;

[0017] The one-point-one-segment database is used to implement the logic of converting candidates' scores and rankings. After structured modeling, it supports fast mapping of any score to ranking and ranking to score, and supports advanced computing capabilities such as interpolation, cumulative ranking, and multi-year comparison.

[0018] The subject selection requirements database covers the specific restrictions on first-choice and second-choice subjects for various majors in various universities under the new college entrance examination system. The data sources of this database include multiple channels such as the Ministry of Education's Sunshine College Entrance Examination Platform, the Admissions and Examination Information Network, and university official websites. After standardization of format and logical modeling, it realizes key functions such as automatic matching with users' subject selection combinations, screening of eligible majors, elimination of incompatible majors, and subject selection risk warnings.

[0019] Preferably, the data construction and analysis module uses a Python-written web crawler program during the development phase to extract data such as college entrance examination ranking tables, subject selection requirements tables, and admission brochures from authoritative data sources such as the Admissions and Examination Information Network and the Sunshine College Entrance Examination Platform.

[0020] After the data is crawled, the pandas, openpyxl, and re tool libraries are used to complete the steps of field standardization, null value handling, logical verification, duplicate value removal, and unit unification, generating a highly consistent and readable structured data file. Semantic rule modeling is also performed on the subject selection requirements, that is, the text description is transformed into a computable Boolean expression.

[0021] The final output data is uploaded to the knowledge base module and used with a code translator to complete format adaptation and test loading. The code structure of this module has reserved update interfaces, which can support automatic re-capture of tasks daily, weekly and monthly, as well as manual version updates, to achieve knowledge timeliness and version control.

[0022] Preferably, the semantic interaction configuration module sets up a multi-layered Prompt in the Prompt editor of the Qianfan platform, including system instructions, user instructions templates, memory prompts, and behavioral guidelines.

[0023] The system commands embed role identity, system capability boundaries, and priority task sequences; the user command templates contain built-in parameter variables to ensure consistency in input format; the behavior guidelines set answer length, citation style, politeness level, and output structure rules to ensure that the system's answer style is professional and easy to read.

[0024] Preferably, the answer synthesis and feedback module is responsible for organizing the university and major recommendation results into tables, segmented text, or graphic formats, and adding recommendation reasons, matching explanations, and precautions to enhance the interpretability and decision transparency of the system. In semantic question answering, this module automatically compresses the document retrieval summary into a natural language answer and inserts reference links and data sources in appropriate locations. It also supports a user feedback mechanism, which can automatically ask whether the user is satisfied with the recommendation after the system outputs the results, guiding the user to make fine adjustments or engage in multiple rounds of interaction to continuously iterate the recommendation results.

[0025] Compared with the prior art, the beneficial effects of the present invention are:

[0026] The system in this application is built using Baidu Cloud Qianfan Platform App Builder to create a college entrance examination application AIAgent. It integrates a custom knowledge base and various college entrance examination plugin tools, combined with large model reasoning capabilities, to create a knowledge-driven, semantic question answering, and intelligent recommendation application assistance platform. This platform aims to structurally integrate complex policy requirements and college information, reduce the understanding cost for candidates and parents, improve the scientific nature and efficiency of application, and has good scalability, making it compatible with other provinces in China that have implemented the new college entrance examination system.

[0027] Intelligent recommendation based on scores and subject selection criteria: Based on the user's input of college entrance examination scores, subject selection combinations, and regional preferences, combined with historical score distribution data and subject selection requirements in the knowledge base, a list of recommended colleges and majors that can be applied for is dynamically generated, supporting multi-dimensional filtering;

[0028] Real-time query of college and major information: With the help of Baidu College Entrance Examination and Baidu AI search plugin, the system supports users to query information such as college overview, major training program, and enrollment data in real time, so as to achieve comprehensive supplementation of content and assist in judgment;

[0029] Supports conversational intelligent question answering and multi-round decision-making interaction: Users can freely ask questions and communicate with the AI ​​Agent, and the system will provide accurate answers based on semantic understanding and knowledge base structure, with a certain ability to logical reasoning and context preservation;

[0030] To ensure the accuracy and interpretability of the recommendations: Each recommendation will be accompanied by a data source and logical explanation to help users understand the basis for the recommendation and enhance the transparency and credibility of the system. Attached Figure Description

[0031] Figure 1 This is a diagram showing the module composition of this system. Detailed Implementation

[0032] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0033] Example:

[0034] Please see Figure 1 The present invention provides a technical solution:

[0035] A college entrance examination application assistance system includes a user input module, a plug-in calling module, a knowledge base system, a data construction and analysis module, a semantic interaction configuration module, and an answer synthesis and feedback module;

[0036] The user input module is the entry point for all interactions. Its core task is to receive the user's initial input, including but not limited to college entrance examination scores, subject combinations, regional and university preferences, and keywords related to major intentions. It supports a combination of free text and structured input. The system uses Prompt engineering and contextual understanding to parse natural language into parameterized information and dynamically fill it into the semantic request template of the model.

[0037] The plugin calling module is responsible for expanding the model's knowledge boundaries and data retrieval capabilities. It integrates the native plugin capabilities provided by Baidu Cloud Qianfan, currently calling three types of plugins: Baidu College Entrance Examination Plugin, Baidu AI Search Plugin, and Code Translator Plugin. The Baidu College Entrance Examination Plugin provides real-time enrollment policies, university databases, and batch information; the Baidu AI Search Plugin supports semantic retrieval of policy and professional Q&A; the Code Translator executes Python scripts to achieve data collection and cleaning; the knowledge base tool loads structured one-point-segment data and subject selection rule tables; and custom datasets supplement unofficial professional databases and admission probability estimation logic.

[0038] The knowledge base system undertakes the basic support work for the system's interpretability, reasonability, and high-precision recommendation tasks. The system consists of two core sub-bases.

[0039] The knowledge base is uploaded and hosted in structured JSON and tabular formats within the knowledge component built into the Qianfan platform. Models can be accessed through various methods such as semantic search and key-value pair lookup. Its design meets the needs of cross-regional migration (it can be expanded to other provinces simply by changing the data), and it has good versatility and maintenance efficiency.

[0040] The data construction and analysis module is the starting point for building the system's knowledge source. It is responsible for the programmatic collection, structured cleaning and model reorganization of external public data resources, as well as the supply of high-quality training data.

[0041] The semantic interaction configuration module is located upstream in the large language model call chain and is responsible for imposing instruction constraints and style training on the model.

[0042] The response synthesis and feedback module is the intersection of all model inputs and knowledge retrieval. It receives the main inference results of the model, the feedback data from the plugins, and the matching results from the knowledge base. It then refines the language, standardizes the format, and organizes the logic to ultimately form a structured and understandable output.

[0043] The system supports setting dynamic opening remarks. When a user enters the system for the first time, the model will automatically generate a background message based on the time, region, and user role, such as "Hello, I am the college application AI assistant. Welcome to use this system for intelligent college application planning. Please tell me your college entrance examination scores and subject combinations, and I will recommend suitable universities and majors for you." This mechanism greatly enhances user trust and the system's human-like experience.

[0044] The user input module also integrates user role recognition. The system determines whether a user is a student, parent, or teacher / advisor based on their language style, and then automatically adjusts the dialogue style and recommendation method accordingly. When a user enters the system for the first time, they will receive an initial opening message to guide them in inputting key parameters, reducing the operational threshold and enhancing the perception of personalized services. The user input module is not only the entry point for input but also the starting point for maintaining the context. The model maintains the user's input state through multi-turn semantic association, achieving state maintenance and task tracking in long dialogue scenarios.

[0045] In the plugin calling module, the Baidu College Entrance Examination plugin is used to query authoritative information on universities, majors, historical admission data, and educational levels; the Baidu AI Search plugin is used for advanced semantic question answering tasks, such as retrieving unstructured documents like admission brochures, professional training programs, employment trends, and national policies, and the model performs language summarization and answers; the code translator plugin is used to execute custom data processing scripts, such as web crawling, data cleaning, field alignment, and data structure transformation, enabling the Agent to evolve and adjust the knowledge base structure and content as needed during runtime.

[0046] The plugin module is tightly integrated with the model invocation logic and has the ability to make multi-round collaborative calls. That is, when the model fails to obtain the answer from the knowledge base when the user asks a question, the model will automatically choose whether to call the plugin and integrate the returned results.

[0047] The knowledge base system consists of two core sub-bases: a score-segment database and a subject selection requirements database.

[0048] The one-point-one-segment database is used to implement the logic of converting candidates' scores and rankings. After structured modeling, it supports fast mapping of any score to ranking and ranking to score, and supports advanced computing capabilities such as interpolation, cumulative ranking, and multi-year comparison.

[0049] The subject selection requirements database covers the specific restrictions on first-choice and second-choice subjects for various majors in various universities under the new college entrance examination system. The data sources of this database include multiple channels such as the Ministry of Education's Sunshine College Entrance Examination Platform, the Admissions and Examination Information Network, and university official websites. After standardization of format and logical modeling, it realizes key functions such as automatic matching with users' subject selection combinations, screening of eligible majors, elimination of incompatible majors, and subject selection risk warnings.

[0050] The data construction and analysis module was developed using a Python-written web crawler to extract data such as college entrance examination ranking tables, subject selection requirements tables, and admission brochures from authoritative data sources such as the Admissions and Examination Information Network and the Sunshine College Entrance Examination Platform.

[0051] After the data is crawled, the pandas, openpyxl, and re tool libraries are used to complete the steps of field standardization, null value handling, logical verification, duplicate value removal, and unit unification, generating a highly consistent and readable structured data file. Semantic rule modeling is also performed on the subject selection requirements, that is, the text description is transformed into a computable Boolean expression.

[0052] The final output data is uploaded to the knowledge base module and used with a code translator to complete format adaptation and test loading. The code structure of this module has reserved update interfaces, which can support automatic re-capture of tasks daily, weekly and monthly, as well as manual version updates, to achieve knowledge timeliness and version control.

[0053] The semantic interaction configuration module sets up a multi-layered Prompt in the Prompt editor of the Qianfan platform, including system instructions, user instructions templates, context review prompts, and behavioral guidelines.

[0054] The system commands embed role identity, system capability boundaries, and priority task sequences; the user command templates contain built-in parameter variables to ensure consistency in input format; the behavior guidelines set answer length, citation style, politeness level, and output structure rules to ensure that the system's answer style is professional and easy to read.

[0055] The answer synthesis and feedback module is responsible for organizing the university and major recommendation results into tables, segmented text, or graphic formats, and adding recommendation reasons, matching status explanations, and precautions to enhance the interpretability and decision transparency of the system. In semantic question answering, this module automatically compresses the document retrieval summary into a natural language answer and inserts reference links and data sources in appropriate locations. It also supports a user feedback mechanism, which can automatically ask whether the user is satisfied with the recommendation after the system outputs the results, guiding the user to make fine adjustments or engage in multiple rounds of interaction to continuously iterate the recommendation results.

[0056] The specific implementation steps of this system are as follows:

[0057] Score Input and Matching Recommendations: After the user inputs their college entrance examination scores, subject combinations, and preferred admission batch, the system returns matching universities and majors based on the knowledge base.

[0058] Input parameters:

[0059] Total score in the college entrance examination (e.g., 546 points), subject combination (e.g., physics + chemistry + biology), interests and hobbies;

[0060] System processing logic:

[0061] After parsing the user input, the matching process begins; the score distribution table is queried to determine the current score's ranking within the province; eligible majors are filtered based on subject selection requirements; and admission data from the past three years is compared to estimate the probability of admission.

[0062] Output content:

[0063] List of matching universities (sorted by recommendation level), information on matching majors (including admission scores and subject selection matching degree), and estimated probability of admission (A: high probability; B: medium probability; C: low probability).

[0064] Intelligent Q&A for Documents:

[0065] Leveraging Baidu's AI search plugin, users can ask questions in natural language about major introductions, admission policies, and employment prospects. The AI ​​Agent calls the document query interface to perform semantic matching and outputs a summary of the answer.

[0066] Example interaction:

[0067] "What jobs can I get after graduating with a finance degree?"; "How do I fill out my college application for Inner Mongolia in 2025?"

[0068] Return structure:

[0069] Document summary (within 300 words); source link (such as the original policy text or school introduction); this module emphasizes the combination of retrieval and generation, providing policy authority and semantic interpretation capabilities;

[0070] Analysis of ranking in one segment:

[0071] It reads the "Inner Mongolia 2025 Score Distribution Table" and supports candidate score location, interval statistics, and cumulative ranking calculation.

[0072] Functional points:

[0073] When a user enters a score, the system returns the score's ranking within the province; when a user queries a range, the system returns the number of people in that score range.

[0074] Example query:

[0075] Input: 546 points (liberal arts); Output: Number of people in this range: 249, cumulative ranking: 42011; This data comes from a knowledge base file crawled and structured by a web crawler.

[0076] University / Major Subject Matching: Matching the compatibility between the candidate's subject combination and the subject requirements of the target major. A structured cleaning logic was designed to address differences in expression types such as compatibility, multiple selections, and "physics or history + no restrictions".

[0077] Example data structure:

[0078] {"Major":"Finance";"Subject Requirements":"Physics or History + No Restriction";"University":"Inner Mongolia University";"Recommendation Level":"Compatible"}; Complex logic parsing is achieved through syntax tree and regular expression processing (see the cleaning module code for details).

[0079] The system uses the following crawler class (InnerMongoliaGaokaoSpider) to crawl the following three types of data: Inner Mongolia's historical score lines (2022–2024); the 2024 arts and science score distribution table (approximately 1000 records); and the 2025 new college entrance examination subject selection requirements (covering five categories).

[0080] The main class structure of web crawlers is as follows:

[0081]

[0082] In actual deployment, the Baidu Qianfan platform code translator is used to run the main function of this class:

[0083] if __name__ == "__main__":

[0084] spider=InnerMongoliaGaokaoSpider()

[0085] spider.run()

[0086] The results will be saved to the / data / directory for use in knowledge base loading and subsequent cleaning.

[0087] To ensure that the AI ​​Agent can accurately identify and retrieve structured data, the system is designed with a data cleaning module, GaokaoDataCleaner.py, and the process is as follows:

[0088] Module Description

[0089]

[0090] Code implementation snippet (using clean_score_ranking as an example)

[0091]

[0092]

[0093] The other two functions rename fields, clean up null values, and normalize the score line table and subject selection requirement table. Finally, the three types of data are saved in CSV format in the / data / folder for import into the knowledge base.

[0094] The system is designed with multiple semantic interaction templates (Prompts) to adapt to different user scenarios, including:

[0095] Role Instructions: You are a college application consultant familiar with Inner Mongolia's college entrance examination policies;

[0096] Background: The student is currently filling out college applications for 2025 and is located in Inner Mongolia Autonomous Region.

[0097] Customized opening remarks: Welcome to the new college entrance examination application system. I am your intelligent application planner. Based on AI big data model technology and policy and education data, I can provide you with precise and personalized college application planning services.

[0098] Each type of question (such as "Which schools can I apply to with a score of 546?") can be matched with a specific interaction intent.

[0099] The Prompt can be edited in the Agent configuration interface of App Builder, and supports multi-turn conversation state maintenance;

[0100] The dependencies between system modules are as follows:

[0101]

[0102] This system aims to deliver efficient services to end users with minimal system resource costs and optimized response speed using the App Builder tool on the Baidu Cloud Qianfan Platform. Deployment objectives include, but are not limited to:

[0103] Rapid deployment: Enables minute-level releases, reducing time-consuming steps such as version compilation, packaging, and configuration in the traditional software release process;

[0104] Flexible expansion: It supports flexible expansion of call capabilities within the platform based on the number of users and usage scenarios, and has good scalability.

[0105] Security and compliance: The entire deployment process and service access support HTTPS encrypted transmission to ensure the security of user data and system interface calls;

[0106] Easy to maintain: The deployment architecture has good monitorability and traceability, which facilitates subsequent operation and maintenance operations such as log analysis, version upgrades and data repair;

[0107] The system will ultimately be integrated and deployed using the multi-module structure provided by the Qianfan platform, including front-end page access, Agent entity construction, knowledge base binding, code tool embedding, dataset management, plugin capability invocation and release strategy configuration, etc.

[0108] The system deployment platform is Baidu Cloud Qianfan, which has the following advantages: natively adapted to Baidu's large model system (deepseek-R1 series); provides App Builder graphical configuration and script hybrid capabilities; built-in plugin market (Baidu College Entrance Examination, Baidu AI Search plugin, etc.); supports private knowledge base data upload; supports multi-round conversation control and role setting configuration; therefore, this project no longer relies on independently deployed servers or self-hosted large model inference services, but instead relies on the platform's "serverless" architecture to build intelligent services.

[0109] Before system deployment, ensure that the following three types of data sources are fully prepared and uploaded to the platform:

[0110] Structured data format requirements: Structured data should be in UTF-8 encoded Excel spreadsheet format, supporting both .xlsx and .xls extensions. Field names must be explicit, and the number of columns should be controlled between 5 and 15 to avoid platform recognition errors.

[0111] Uploaded file requirements: admission_scores.xlsx: containing the fields "Year, Subject Category, Batch, Cutoff Score, Source".

[0112] score_ranking.xlsx: contains the fields "score, subject category, number of students in this range, cumulative number of students, and year";

[0113] subject_requirements.xlsx: contains fields such as "major name, subject selection requirements, institution type, major category, province, and applicable year";

[0114] The platform automatically parses the uploaded Excel file and identifies the data based on the column names in the first row of each field.

[0115] Data upload steps: On the App Builder page:

[0116] 1. Select either the "Knowledge Base" or "Tool Configuration" module;

[0117] 2. Click "Upload New Dataset";

[0118] 3. Upload the .xlsx file and wait for the system to parse the fields;

[0119] 4. Bind the data after confirming that the fields are correctly identified.

[0120] 5. Name the dataset, for example, "2025_Inner Mongolia_Segmented Data", "Inner Mongolia 2025 Undergraduate (Associate) Degree Subject Selection Requirements", etc.;

[0121] Plugin integration and deployment methods:

[0122] Baidu College Entrance Examination Plugin: The Baidu College Entrance Examination Plugin is one of the main query interfaces of this system, and its functions include: university / major search;

[0123] The program includes information on majors, historical admission scores, and rankings; references to subject evaluation results; steps to enable the plugin in Agent Builder; opening the Agent configuration page; adding the plugin module → searching for "Baidu Gaokao"; clicking "Enable" and granting read / call permissions; optional: binding query fields (e.g., entering a score → automatically redirecting to the query); the plugin's response content can be called by the model to generate a comprehensive judgment answer;

[0124] Baidu AI Search Plugin: The Baidu AI Search Plugin is used to query documents such as college admission brochures, employment trend analyses, and regional policies. It is activated in the same way as the Baidu College Entrance Examination Plugin, and can be directly triggered by binding keywords. For example: a user enters: "Introduce the future development direction of software engineering major"; the Agent automatically triggers the Baidu AI Search Plugin to search for "software engineering employment trends" and analyze the results.

[0125] Prompt Configuration and Semantic Strategy Deployment: As the core part of semantic regulation, the deployment process of Prompt configuration includes: writing role identity settings and behavior guidelines in the "System Prompt" area of ​​App Builder, setting welcome messages, response tone, usage logic, rejection logic, etc.

[0126] The core task of this system is to design a precise and personalized college application form for each student based on their gender, interests, subject selection, college entrance examination scores, and future development needs.

[0127] I. Job Preparation: Familiarity with Inner Mongolia's college entrance examination policies, admission rules, and basic information about universities and majors is required. Additionally, proficiency in using tools such as code translators and Baidu AI search is necessary to quickly obtain and filter relevant information.

[0128] II. Role Workflow:

[0129] 1. Student Information Analysis: First, analyze the student's gender, interests, subject selection, and college entrance examination scores to clarify their application direction and positioning;

[0130] 2. University and Major Selection: Based on students' subject selection and grades, the system uses a knowledge base and Baidu AI search to select eligible universities and majors.

[0131] 3. Application Strategy Formulation: Based on students' needs and goals, formulate application strategies of "ambitious" (add 20 points), "stable" (add or subtract 5 points), and "safe" (subtract 20 points), and determine the specific universities and majors under each strategy;

[0132] 4. Ranking of Applications: Based on the application strategy, the recommended universities and majors are ranked.

[0133] 5. Special Requirements Check: Verify whether candidates meet the requirements for the National Free Teacher Training Program, Medical Targeted Program, or Rural Household Special Program, and adjust the recommendation list accordingly;

[0134] 6. Handling of admissions by broad category and by subject: Provide corresponding recommendations and suggestions for admissions by broad category and by subject.

[0135] 7. Final application form preparation: Organize the results of the above steps into a tabular application form, ensuring it is clear and easy to read.

[0136] Explanation of terms and abbreviations:

[0137]

[0138]

[0139] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or basic characteristics. Therefore, the embodiments should be considered exemplary and non-limiting in all respects. The scope of the invention is defined by the appended claims rather than the foregoing description. Therefore, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention, and no reference numerals in the claims should be construed as limiting the scope of the claims.

[0140] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A college entrance examination application assistance system, characterized in that, It includes a user input module, a plugin calling module, a knowledge base system, a data construction and analysis module, a semantic interaction configuration module, and an answer synthesis and feedback module; The user input module is the entry point for all interactions. Its core task is to receive the user's initial input, including but not limited to college entrance examination scores, subject combinations, regional and university preferences, and keywords related to major intentions. It supports a combination of free text and structured input. The system uses Prompt engineering and contextual understanding to parse natural language into parameterized information and dynamically fill it into the semantic request template of the model. The plugin calling module is responsible for expanding the knowledge boundary and data retrieval capabilities of the model. It integrates the native plugin capabilities provided by Baidu Cloud Qianfan and currently calls three types of plugins: Baidu College Entrance Examination Plugin, Baidu AI Search Plugin, and Code Translator Plugin. The knowledge base system undertakes the basic support work for the system's interpretability, reasonability, and high-precision recommendation tasks. The system consists of two core sub-bases. The data construction and analysis module is the starting point for building the system's knowledge source. It is responsible for the programmatic collection, structured cleaning and model reorganization of external public data resources, as well as the supply of high-quality training data. The semantic interaction configuration module is located upstream in the large language model call chain and is responsible for imposing instruction constraints and style training on the model. The response synthesis and feedback module is the intersection of all model inputs and knowledge retrieval. It receives the main inference results of the model, the feedback data from the plugins, and the matching results from the knowledge base. It then refines the language, standardizes the format, and organizes the logic to ultimately form a structured and understandable output.

2. The college entrance examination application assistance system according to claim 1, characterized in that: The user input module also integrates user role recognition. The system determines whether a user is a student, parent, or teacher / advisor based on their language style, and then automatically adjusts the dialogue style and recommendation method accordingly. When a user enters the system for the first time, they will receive an initial opening message to guide them in inputting key parameters, reducing the operational threshold and enhancing the perception of personalized services. The user input module is not only the entry point for input but also the starting point for maintaining the context. The model maintains the user's input state through multi-turn semantic association, achieving state maintenance and task tracking in long dialogue scenarios.

3. The college entrance examination application assistance system according to claim 1, characterized in that: In the plugin calling module, the Baidu College Entrance Examination plugin is used to query authoritative information on universities, majors, historical admission data, and educational levels; the Baidu AI Search plugin is used for advanced semantic question answering tasks, such as retrieving unstructured documents like admission brochures, professional training programs, employment trends, and national policies, and the model performs language summarization and answers; the code translator plugin is used to execute custom data processing scripts, such as web crawling, data cleaning, field alignment, and data structure transformation, enabling the Agent to evolve and adjust the knowledge base structure and content as needed during runtime. The plugin module is tightly integrated with the model invocation logic and has the ability to make multi-round collaborative calls. That is, when the model fails to obtain the answer from the knowledge base when the user asks a question, the model will automatically choose whether to call the plugin and integrate the returned results.

4. The college entrance examination application assistance system according to claim 1, characterized in that: The knowledge base system consists of two core sub-bases: a score-segment database and a subject selection requirements database. The one-point-one-segment database is used to implement the logic of converting candidates' scores and rankings. After structured modeling, it supports fast mapping of any score to ranking and ranking to score, and supports advanced computing capabilities such as interpolation, cumulative ranking, and multi-year comparison. The subject selection requirements database covers the specific restrictions on first-choice and second-choice subjects for various majors in various universities under the new college entrance examination system. The data sources of this database include multiple channels such as the Ministry of Education's Sunshine College Entrance Examination Platform, the Admissions and Examination Information Network, and university official websites. After standardization of format and logical modeling, it realizes key functions such as automatic matching with users' subject selection combinations, screening of eligible majors, elimination of incompatible majors, and subject selection risk warnings.

5. The college entrance examination application assistance system according to claim 1, characterized in that: The data construction and analysis module was developed using a Python-written web crawler to extract data such as college entrance examination ranking tables, subject selection requirements tables, and admission brochures from authoritative data sources such as the Admissions and Examination Information Network and the Sunshine College Entrance Examination Platform. After the data is crawled, the pandas, openpyxl, and re tool libraries are used to complete the steps of field standardization, null value handling, logical verification, duplicate value removal, and unit unification, generating a highly consistent and readable structured data file. Semantic rule modeling is also performed on the subject selection requirements, that is, the text description is transformed into a computable Boolean expression. The final output data is uploaded to the knowledge base module and used with a code translator to complete format adaptation and test loading. The code structure of this module has reserved update interfaces, which can support automatic re-capture of tasks daily, weekly and monthly, as well as manual version updates, to achieve knowledge timeliness and version control.

6. The college entrance examination application assistance system according to claim 1, characterized in that: The semantic interaction configuration module sets up a multi-layered Prompt in the Prompt editor of the Qianfan platform, including system instructions (SystemPrompt), user instruction templates (User Prompt), context review prompts (Memory Prompt), and behavioral specification guidance (Behavioral Prompt). The system commands embed role identity, system capability boundaries, and priority task sequences; the user command templates contain built-in parameter variables to ensure consistency in input format; the behavior guidelines set answer length, citation style, politeness level, and output structure rules to ensure that the system's answer style is professional and easy to read.

7. The college entrance examination application assistance system according to claim 1, characterized in that: The answer synthesis and feedback module is responsible for organizing the university and major recommendation results into tables, segmented text, or graphic formats, and adding recommendation reasons, matching status explanations, and precautions to enhance the interpretability and decision transparency of the system. In semantic question answering, this module automatically compresses the document retrieval summary into a natural language answer and inserts reference links and data sources in appropriate locations. It also supports a user feedback mechanism, which can automatically ask whether the user is satisfied with the recommendation after the system outputs the results, guiding the user to make fine adjustments or engage in multiple rounds of interaction to continuously iterate the recommendation results.

Images