College entrance examination volunteer filling assisting method and system based on large language model

Abstract

The invention provides a college entrance examination volunteer filling assisting method and system based on a large language model, and the method is characterized in that the method comprises the steps: presetting a question template and an SQL template according to the volunteer filling scene demands and a database schema; analyzing a natural language query Q by adopting a large language model and designing a cue word, and combining a question template sequence number and an analysis result to generate a structured query condition set CQ; generating a database SQL query statement according to the CQ in combination with a preset SQL template; and managing database connection by using a connection pool technology, executing SQL query, generating a result set RS, and outputting information to a user. Through the method, the query efficiency and the interaction friendliness of the volunteer filling auxiliary system are improved.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and particularly to an auxiliary method and system for college entrance examination volunteer filling based on a large language model. Background Art

[0002] The college entrance examination (National College Entrance Examination) is a crucial part of the Chinese education system, with millions of candidates taking it every year. The college entrance examination scores are one of the key factors determining whether candidates can be admitted to their desired universities. An efficient and accurate college entrance examination score query system is of great significance to candidates, parents, and educational institutions.

[0003] Currently, the mainstream college entrance examination score query products mainly rely on traditional database query technologies and simple user interface designs. These products usually retrieve data from the database through SQL statements and display the query results to users through web pages or mobile applications. When querying, users need to manually input personal information, such as province, subject category, school, etc. The system executes corresponding SQL queries based on the input information and returns the matching results. For example, in the Chinese invention application "A Method and Device for Intelligent Recommendation of College Entrance Examination Volunteers Based on Big Data" with the publication number CN104239499A, by inputting detailed classification query information, it provides intelligent college entrance examination volunteer recommendations to assist in applying and improve the success rate of application. However, this method is cumbersome to operate, has limitations in user input, and has poor user-friendliness. When users use it, the query process is complex, with multiple page switches and a poor experience. In addition, these products mainly adopt traditional SQL query methods and fail to fully utilize natural language processing technologies to improve query efficiency and user experience, having obvious technical limitations.

[0004] SQL (Structured Query Language) is a standard language for managing and querying relational databases. In the volunteer filling service, SQL queries are widely used to retrieve the admission cut-off scores, major settings, and admission trends over the years of specific schools from the college enrollment database to assist students and parents in making wise volunteer choices. Traditional database query systems mainly rely on SQL statements for data retrieval. Although they can provide accurate query results, they require users to have certain technical knowledge and have a relatively high operation complexity. Most existing college entrance examination volunteer filling service platforms adopt this mode, where users need to manually input query conditions, and the system returns results by executing SQL statements. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to improve the query efficiency and interaction friendliness of the volunteer filling auxiliary system.

[0006] The present invention achieves the solution to the above technical problem through the following technical means:

[0007] The present invention provides a method for assisting in college entrance examination volunteer filling based on a large language model, including:

[0008] S1. Preset question templates and SQL templates according to the requirements of the volunteer filling scenario and the database scheme;

[0009] S2. Use a large language model and design prompt words to parse the natural language query Q, and combine the question template numbers and the parsing results to generate a structured query condition set CQ;

[0010] S3. Generate a database SQL query statement according to CQ in combination with the preset SQL template;

[0011] S4. Use the connection pool technology to manage the database connection, execute the SQL query and generate a result set RS, and output information to the user.

[0012] Further, the step S1 specifically includes:

[0013] (1) Investigate the volunteer filling requirements and user requirement data;

[0014] (2) Based on the investigation data, preset question templates and SQL templates for the scheme of the relational database DB;

[0015] (3) The question templates cover common volunteer filling scenario requirements, the attribute values of the SQL templates are filled with placeholders, and the numbers of the question templates and the SQL templates correspond one by one.

[0016] Further, the step S2 includes the following steps:

[0017] S21. Use a large language model to judge whether it can parse the natural language query request Q initiated by the user. If it holds, continue to execute step S22; otherwise, feedback to the user, request to re-enter the query request, and return to step S21;

[0018] S22. Use a large language model and design prompt words to parse the natural language query Q, identify the user's intention, and judge whether there is a question template in the question template library that matches the user's intention; if it holds, select this question template; otherwise, randomly select a question template;

[0019] S23. Use a large language model to extract the entity information in Q, and organize the number of the selected question template and the extracted entity information into a structured query condition set CQ.

[0020] Further, the step S3 includes the following steps:

[0021] S31. Select the preset SQL template according to the question template number in the query condition set CQ;

[0022] S32. Determine whether the content in the current query condition set CQ can completely replace the placeholder in the SQL template. If it holds, continue to execute step S33; otherwise, request the missing information from the user until it meets the condition of being able to completely replace the placeholder in the SQL template, and then continue to execute step S33;

[0023] S33. Replace all variable values in the SQL template; generate an SQL query statement.

[0024] Further, the step S4 includes the following steps:

[0025] S41. Connect to the database NeuralDB through SQLite3;

[0026] S42. Determine whether the connection is successful. If it holds, continue to execute step S43; otherwise, execute step S45;

[0027] S43. Execute the SQL query and determine whether the query is successful. If it holds, continue to execute step S44; otherwise, execute step S45;

[0028] S44. Obtain the query result, convert the query result into a Markdown table format, and output it to the user;

[0029] S45. Record the error log and return a prompt message.

[0030] The present invention also provides a system for assisting in college entrance examination volunteer filling based on a large language model. When the system runs, it applies the above method and includes the following modules:

[0031] A template creation module for presetting a question template and an SQL template according to the requirements of the volunteer filling scenario and the database scheme;

[0032] A query condition set construction module for using a large language model and designing prompt words to parse the natural language query Q, and combining the question template number and the parsing result to generate a structured query condition set CQ;

[0033] A query statement construction module for generating a database SQL query statement according to CQ in combination with a preset SQL template;

[0034] An output module for using a connection pool technology to manage database connections, execute SQL queries, generate a result set RS, and output information to the user.

[0035] The advantages of the present invention are:

[0036] (1) The query parsing method for college entrance examination volunteer application service based on large language model of the present invention optimizes the problem-solving ability of the general large language model for volunteer application tasks by designing prompt words in the field of college entrance examination volunteer application service. By using the large language model to parse natural language queries, extract information, and identify user intentions, the query efficiency is improved.

[0037] (2) In the field of college entrance examination volunteer application service, this patent organizes the parsed key information into a structured set of query conditions, supports users to input queries in natural language, increases the freedom of users to input questions, and optimizes the usage experience and user-friendliness.

[0038] (3) Under the constraints of the college volunteer application scenario, the present invention presets double templates of problem templates and SQL templates for the database scheme to ensure the accuracy of the output information and provide reliable volunteer application services for users.

[0039] (4) In the field of college entrance examination volunteer application service, the present invention realizes the application of natural language processing technology in the college entrance examination volunteer application service scenario through systematic modular design, including natural language parsing module, SQL generation module, and database query module. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 It is a schematic flow chart of the method for assisting college entrance examination volunteer application based on large language model of the present invention;

[0041] Figure 2 It is a schematic flow chart of step S2 of the present invention;

[0042] Figure 3 It is a schematic flow chart of step S3 of the present invention;

[0043] Figure 4 It is a schematic flow chart of step S4 of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0044] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0045] Embodiment 1

[0046] The college entrance examination score line query system based on natural language processing is a software system that integrates natural language processing (NLP), machine learning (ML), and database management technologies, providing users with an automated service for matching college entrance examination scores with college admission score lines. The system receives users' query requests, parses the key information in the requests, performs database query operations, and presents the results in a user-friendly manner.

[0047] The system can be formally defined as follows: The input is the user's query request Q, which contains the following fields: place of origin S, category C, school U, major M, score Sc, and the number of results to display N; the output is the query result set RS, which contains the following fields: school name UN, category UC, major name MN, minimum admission score SL, link L, etc. The system includes a relational database DB that stores the college admission score lines and related information over the years.

[0048] This embodiment provides a method for assisting in filling out college entrance examination application forms based on large language models, as Figure 1 shown, and the specific steps include:

[0049] S1. Preset question templates and SQL templates according to the requirements of the college entrance examination application form filling scenario and the database scheme. The specific steps include:

[0050] (1) Conduct research on the requirements of college entrance examination application form filling and user requirement data;

[0051] (2) Based on the research data, preset question templates and SQL templates for the scheme of the relational database DB;

[0052] (3) The question templates cover common requirements of college entrance examination application form filling scenarios, and the attribute values of the SQL templates are filled with placeholders. The question templates and SQL templates have a one-to-one correspondence in number.

[0053] For example, the user's requirement is: Obtain the information on the college admission score lines of a certain major in a certain college over the years. The schema of the database gaokao_scores includes the following fields:

[0054] School name: Stores the name of the school, using the variable-length string type (VARCHAR(255)).

[0055] School location: Stores the city where the school is located, using the variable-length string type (VARCHAR(255)).

[0056] Year: Stores the year of the admission data, using the integer type (INT).

[0057] Place of origin: Stores the place of origin information of the students, using the variable-length string type (VARCHAR(100)).

[0058] Category: Stores the category information of students (such as science, liberal arts), using variable-length string type (VARCHAR(50)).

[0059] Major Name: Stores the name of the admitted major, using variable-length string type (VARCHAR(255)).

[0060] Minimum Admission Score: Stores the minimum score for admission to this major, using floating-point type (FLOAT).

[0061] url: Stores the link to relevant information, using variable-length string type (VARCHAR(500)).

[0062] Remarks: Stores additional remarks information, using text type (TEXT).

[0063] Question Template 1: Please provide your place of origin + category + school + major, and inquire about the required scores for xx college xx major you applied for;

[0064] Question Template 2: Please provide your place of origin + category + major + intended city, and inquire about the admission scores of this major in different colleges in xx province / city. For example: Anhui + science + computer + Hefei;

[0065] SQL Template 1: {SELECT*FROM gaokao_scores WHERE place of origin = \"%{shengfen}%\" AND category = \"%{kelei}%\" AND school name = \"%{xuexiao}%\" AND major name = \"%{zhuanye}%\"};

[0066] SQL Template 2: {SELECT*FROM gaokao_scores WHERE place of origin = \"%{shengfen}%\" AND category = \"%{kelei}%\" AND major name = \"%{zhuanye}%\" AND school location = \"%{chengshi}%\"}

[0067] S2. Use a large language model and design prompt words to parse the natural language query Q, and combine the question template number and the parsing result to generate a structured query condition set CQ. As Figure 2 shown, the specific implementation steps are as follows:

[0068] S21. Use a large language model to judge whether it can parse the natural language query request Q initiated by the user. If it holds, continue to execute step S22; otherwise, feedback to the user, request to re-enter the query request, and return to step S21;

[0069] S22. Use a large language model and design prompt words to parse the natural language query Q, identify the user's intention, and determine whether there is a question template in the question template library that matches the user's intention. If so, select this question template; otherwise, randomly select a question template.

[0070] S23. Use a large language model to extract the entity information in Q, and organize the serial number of the selected question template and the extracted entity information into a structured query condition set CQ.

[0071] In the above steps, considering the cost, a general large language model is used in this step, and the effect of the model to complete the task is optimized by designing appropriate prompt words. The basically set prompt words are: template number, place of origin, subject category, school, major, score, intended city, type. For example:

[0072] Place of origin: Name of the province, such as Anhui.

[0073] Subject category: Science and engineering, History, Liberal arts, Physics, Comprehensive reform. The subject category is optional. If the subject entered by the user does not exactly match these options and no result is returned, select a parameter that is closest. For example, change "Physics, Chemistry, Biology" to "Physics", "History, Geography, Biology" to "History", change science and engineering to science, physics, and change liberal arts to history, etc., and try multiple times. If not provided, do not set.

[0074] Intended city: The city where the user wants to apply. If both the place of origin and the intended city exist in the user's input and are the same, then assign both the place of origin and the intended city to the province entered by the user. For example, for science students from Jiangsu (place of origin), who scored xx points this year and want to know which schools in Jiangsu (intended city) they can apply to, then student_origin = 'Jiangsu', preferred_city = 'Jiangsu'; for liberal arts candidates from Guangdong (place of origin), who scored xx points this year and want to know which schools in Guangdong (intended city) they can apply to, then student_origin = 'Guangdong', preferred_city = 'Guangdong', and so on.

[0075] If not provided, do not set.

[0076] Type: Special project, Chinese-foreign cooperation, National special project, Local special project, University special project.

[0077] Examples of question templates (template_type) are as follows:

[0078] 1: Please provide your place of origin + subject category + school + major, and consult the scores required for applying to xx major in xx institution.

[0079] 2: Please provide your place of origin + subject category + major + intended city, and inquire about the admission scores of this major in different universities in xx Province / City. For example: Anhui + Science + Computer + Hefei;

[0080] According to the above template, first determine which template the question belongs to and obtain the template number. Then extract the parameters that may be included in the question: place of origin, subject category, school, major, score, intended city, type. And write them in json format after 'Answer'. The example is as follows:

[0081] Question: I'm from Anhui, majoring in Science, and want to study Computer Science at Anhui University. How many scores do I need?

[0082] Answer: {"Question template number":"1","Place of origin":"Anhui","Subject category":"Science","School name":"Anhui University","Major":"Computer"}

[0083] The parsing effect is as follows:

[0084] #Q: {"I'm from Anhui, majoring in Science, and want to study Computer Science at Anhui University. How many scores do I need?"}

[0085] #CQ: {"Question template number":"1","Place of origin":"Anhui","Subject category":"Science","School name":"Anhui University","Major":"Computer"}

[0086] S3. Generate a database SQL query statement according to CQ combined with the preset SQL template. As Figure 3 shown, the specific implementation steps are:

[0087] S31. Select the preset SQL template according to the question template serial number in the query condition set CQ;

[0088] S32. Judge whether the content in the current query condition set CQ can completely replace the placeholder in the SQL template; if it holds, continue to execute step S33; otherwise, ask the user for the missing information until it meets the condition that it can completely replace the placeholder in the SQL template, and then continue to execute step S33; Filling in the volunteer is a very rigorous matter, which is related to the most suitable development direction of the candidates in the future. Therefore, the method of this embodiment strives to be accurate and error-free for the input and feedback of user information.

[0089] S33. Replace all variable values in the SQL template; generate an SQL query statement.

[0090] The generation effect of the SQL query statement is as follows:

[0091] #SQL Template 1: {SELECT * FROM gaokao_scores WHERE place_of_origin = "%{shengfen}%" AND subject_category = "%{kelei}%" AND school_name = "%{xuexiao}%" AND major_name = "%{zhuanye}%"};

[0092] #SQL: {SELECT * FROM gaokao_scores WHERE school_name = 'Anhui University' AND major_name = 'Computer Science'}

[0093] S4. Use connection pool technology to manage database connections, execute SQL queries and generate a result set RS, and output information to the user. As Figure 4 shown, the specific implementation steps are as follows:

[0094] S41. Connect to the database NeuralDB through SQLite3;

[0095] S42. Judge whether the connection is successful; if it is successful, continue to execute step S43; otherwise, execute step S45;

[0096] S43. Execute the SQL query and judge whether the query is successful; if it is successful, continue to execute step S44; otherwise, execute step S45;

[0097] S44. Obtain the query result, convert the query result into Markdown table format and output it to the user;

[0098] S45. Record the error log and return a prompt message.

[0099] An example of the query result is as follows:

[0100] #RS: [{"school_name": "Anhui University", "year": "2023", "place_of_origin": "Anhui Province", "subject_category": "Science and Engineering", "major_name": "Computer Science", "minimum_admission_score": "2023", "url": "https: / / bkzs.ahu.edu.cn / static / front / ahu / basic / html_web / lnfs.html"},...].

[0101] Embodiment 2

[0102] It should be further noted that, based on the same inventive concept, the present invention also provides a college entrance examination volunteer filling assistance system based on a large language model. When the system runs, it executes the method described in Embodiment 1, including the following modules:

[0103] A template creation module for presetting question templates and SQL templates according to the requirements of the college entrance examination volunteer filling scenario and the database scheme. The specific execution method of the template creation module is as follows:

[0104] (1) Investigate the data of college entrance examination volunteer filling requirements and user requirements;

[0105] (2) Based on the investigation data, preset question templates and SQL templates for the scheme of the relational database DB;

[0106] (3) The question templates cover the common requirements of the college entrance examination volunteer filling scenario, and the attribute values of the SQL templates are filled with placeholders. The question templates and SQL templates have corresponding numbers one by one.

[0107] A query condition set construction module for using a large language model and designing prompt words to parse the natural language query Q, and combining the question template number and the parsing result to generate a structured query condition set CQ. The specific execution method of the query condition set construction module is as follows:

[0108] (1) Use a large language model to judge whether it can parse the natural language query request Q initiated by the user. If it holds, continue to execute step S22; otherwise, feedback to the user, request to re-enter the query request, and return to step (1);

[0109] (2) Use a large language model and design prompt words to parse the natural language query Q, perform intention recognition on the user, and judge whether there is a question template in the question template library that matches the user's intention. If it holds, select this question template; otherwise, randomly select a question template;

[0110] (3) Use a large language model to extract the entity information in Q, and organize the number of the selected question template and the extracted entity information into a structured query condition set CQ.

[0111] A query statement construction module for generating a database SQL query statement according to CQ in combination with the preset SQL template. The specific execution method of the query statement construction module is as follows:

[0112] (1) Select the preset SQL template according to the question template number in the query condition set CQ;

[0113] (2) Judge whether the content in the current query condition set CQ can completely replace the placeholders in the SQL template. If it holds, continue to execute step (3); otherwise, ask the user for the missing information until it meets the condition of being able to completely replace the placeholders in the SQL template, and then continue to execute step (3);

[0114] (3) Replace all variable values in the SQL template; generate an SQL query statement.

[0115] An output module, which is used to manage database connections using connection pool technology, execute SQL queries and generate a result set RS, and output information to users. The specific execution method of the output module is as follows:

[0116] (1) Connect to the database NeuralDB through SQLite3;

[0117] (2) Determine whether the connection is successful; if it is successful, continue to execute step (3); otherwise, execute step (5);

[0118] (3) Execute an SQL query and determine whether the query is successful; if it is successful, continue to execute step (4); otherwise, execute step (5);

[0119] (4) Obtain the query result, convert the query result into Markdown table format and output it to the user;

[0120] (5) Record the error log and return a prompt message.

[0121] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for assisting college entrance examination volunteer application based on a large language model, characterized in that: include: S1. Preset question templates and SQL templates according to volunteer application scenario requirements and database scheme; S2. Use a large language model and design prompt words to parse the natural language query Q, and combine the question template number and the parsing result to generate a structured query condition set CQ; S3. Generate a database SQL query statement based on the CQ and the preset SQL template; S4. Use connection pool technology to manage database connections, execute SQL queries and generate result sets RS, and output information to users.

2. The method for assisting college entrance examination volunteer filling based on a large language model according to claim 1, characterized in that: The step S1 specifically includes: (1) Investigate volunteer application needs and user demand data; (2) Based on the survey data, preset question templates and SQL templates for the relational database scheme; (3) The question template covers the common volunteer application scenario requirements. The attribute values ​​of the SQL template are filled with placeholders, and the question template corresponds to the SQL template number one by one.

3. The method for assisting college entrance examination volunteer filling based on a large language model according to claim 1, characterized in that: The step S2 comprises the following steps: S21, using the large language model to determine whether the natural language query request Q initiated by the user can be parsed, if yes, continue to step S22; otherwise, feedback is given to the user to ask him to re-enter the query request, and then return to step S21; S22, using a large language model and designing prompt words to parse the natural language query Q, identify the user's intention, and determine whether there is a question template in the question template library that matches the user's intention; if yes, select this question template; otherwise, randomly select a question template; S23. Use a large language model to extract entity information in Q, and organize the serial number of the selected question template and the extracted entity information into a structured query condition set CQ.

4. The method for assisting college entrance examination volunteer filling based on a large language model according to claim 3 is characterized in that: The step S3 comprises the following steps: S31, selecting a preset SQL template according to the question template sequence number in the query condition set CQ; S32, determine whether the content in the current query condition set CQ can completely replace the placeholder in the SQL template; if so, continue to step S33; otherwise, ask the user for the missing information until the placeholder in the SQL template can be completely replaced, and continue to step S33; S33. Replace all variable values ​​in the SQL template; generate an SQL query statement.

5. The method for assisting college entrance examination volunteer filling based on a large language model according to claim 1, characterized in that: The step S4 comprises the following steps: S41. Connect to the NeuralDB database via SQLite3. S42, determine whether the connection is successful; if so, proceed to step S43; otherwise, proceed to step S45; S43, execute SQL query to determine whether the query is successful; if yes, continue to step S44; otherwise, execute step S45; S44, obtaining the query results, converting the query results into a Markdown table format, and outputting them to the user; S45. Record an error log and return a prompt message.

6. A system for assisting students in filling out college entrance examination applications based on a large language model, characterized in that: When the system is running, the method described in claims 1 to 5 is applied, including the following modules: Template creation module, used to preset question templates and SQL templates according to volunteer application scenario requirements and database scheme; The query condition set building module is used to adopt a large language model and design prompt words to parse the natural language query Q, and combine the question template sequence number and the parsing result to generate a structured query condition set CQ; The query statement building module is used to generate database SQL query statements based on CQ and preset SQL templates; The output module is used to manage database connections using connection pool technology, execute SQL queries and generate result sets RS, and output information to users.

7. The system for assisting college entrance examination volunteers filling out application based on a large language model according to claim 6 is characterized in that: The specific execution mode of the template creation module is as follows: (1) Investigate volunteer application needs and user demand data; (2) Based on the survey data, preset question templates and SQL templates for the relational database scheme; (3) The question template covers the common volunteer application scenario requirements. The attribute values ​​of the SQL template are filled with placeholders, and the question template corresponds to the SQL template number one by one.

8. The system for assisting college entrance examination volunteers filling out application based on a large language model according to claim 6 is characterized in that: The query condition set building module is specifically executed as follows: (1) Using the large language model to determine whether the natural language query request Q initiated by the user can be parsed, if yes, continue to step S22; otherwise, feedback is given to the user to ask him to re-enter the query request, and then return to step (1); (2) Use a large language model and design prompt words to parse the natural language query Q, identify the user's intent, and determine whether there is a question template in the question template library that matches the user's intent; If yes, select this question template; Otherwise, a question template is randomly selected; (3) Use a large language model to extract entity information in Q, and organize the serial number of the selected question template and the extracted entity information into a structured query condition set CQ.

9. The system for assisting college entrance examination volunteers filling out application based on a large language model according to claim 8 is characterized in that: The specific execution mode of the query statement construction module is as follows: (1) Select a preset SQL template according to the question template sequence number in the query condition set CQ; (2) Determine whether the content in the current query condition set CQ can completely replace the placeholder in the SQL template; if so, continue to execute step (3); otherwise, ask the user for the missing information until the placeholder in the SQL template can be completely replaced, and continue to execute step (3); (3) Replace all variable values ​​in the SQL template; generate SQL query statements.

10. The system for assisting college entrance examination volunteers filling out application based on a large language model according to claim 6, characterized in that: The specific implementation mode of the output module is: (1) Connect to the NeuralDB database through SQLite3; (2) Determine whether the connection is successful; if so, proceed to step (3); otherwise, proceed to step (5); (3) Execute the SQL query to determine whether the query is successful; if so, proceed to step (4); otherwise, proceed to step (5); (4) Obtain the query results, convert the query results into Markdown table format, and output them to the user; (5) Record error logs and return prompt information.

Images