Database-based course data online retrieval method

Abstract

The invention discloses a database-based curriculum data online retrieval method, and relates to the technical field of curriculum data online retrieval, and the method comprises the following steps: S1, constructing a curriculum database table which comprises the following fields: curriculum names, curriculum starting time, difficulty levels and curriculum keywords; according to the course data online retrieval method based on the database, an overall technical system is formed through multi-dimensional retrieval conditions, weight sorting and fuzzy extension to intention perception and performance optimization; intelligent and efficient retrieval is realized by combining fuzzy rule extension, user intention perception and multi-layer index optimization; the problems of inaccurate retrieval and slow response speed in a large-scale course data scene are solved, and the user experience is remarkably improved; the defects of an existing method in the process of processing complex requirements are overcome, and the retrieval accuracy and efficiency are remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of online course data retrieval, and in particular to an online course data retrieval method based on a database. Background Art

[0002] With the rapid development of online education platforms, the amount of course data has exploded, covering courses of various types and difficulty levels. Existing retrieval methods are usually based on simple keyword matching or static classification screening, which are difficult to meet the complex needs of users, such as multi-condition dynamic screening by time, difficulty and subject. The problems of low retrieval efficiency and inaccurate results have significantly affected the user experience. Therefore, there is an urgent need for an online retrieval method for course data that can support multi-dimensional screening and weighted sorting. Summary of the invention

[0003] In order to solve the above technical problems, the present invention is implemented by the following technical solutions: a method for online retrieval of course data based on a database, comprising the following steps: S1: Construct a course database table, which includes the following fields: course name, course start time, difficulty level, and course keywords; S2: Define multi-dimensional search conditions, including time dimension, difficulty dimension and subject keyword dimension; S3: Provides a user front-end input interface for selecting or entering search conditions; S4: Based on the search conditions input by the user, dynamically generate database query statements, perform multi-dimensional screening, and obtain screening results; S5: Sort the screening results through a weight sorting algorithm, and give priority to displaying the courses that best match user needs.

[0004] The step S3 includes introducing deep learning intention perception technology into the course retrieval scenario, and the process is as follows: User intent perception model: Based on the keywords, historical behaviors, and contextual conditions entered by the user, the deep learning model BERT is used to predict the user's search intent; Intelligent result recommendation: In the search results, the courses that best meet the user's intention are displayed first. For conditions that the user has not specified, the classified results with different intentions are recommended to obtain the recommended results; Dynamic feedback optimization: Based on the user’s click behavior on the recommended results, the prediction weight of the intent-aware model is adjusted to improve the accuracy of the recommendation.

[0005] The multi-dimensional search conditions in step S2 include: Time dimension: supports users to filter courses by specific time periods or cycles, such as weekends and evenings; Difficulty dimension: filter based on the difficulty level field in the course data, where the difficulty levels include beginner, intermediate, and advanced; Topic keyword dimension: supports users to enter keywords and filter based on full-text matching of course titles, introductions or tags.

[0006] Preferably, the method further includes a step of dynamically adjusting the weight based on user behavior data, the step comprising: Collect user behavior data after searching, including user click frequency, stay time, browsing, collection and registration operations; Build a user preference model based on user behavior data and extract the user's preference weights for different condition dimensions; When the user does not manually set the weights, the weights are dynamically adjusted according to the user preference model.

[0007] Preferably, in step S3, an intelligent condition extension mechanism based on fuzzy rules is introduced, and the intelligent condition extension mechanism includes the following steps: Use the keyword synonym library and word vector model Word2Vec to calculate the semantic similarity of the keywords entered by the user and dynamically expand the keyword field; When the time condition set by the user has no result, the time range is automatically relaxed to a nearby time period; Provide explicit identification of "extended mode" in search results and allow users to choose to enable or disable extended mode.

[0008] Preferably, the keyword synonym library is generated by combining manual construction and automatic learning, specifically including: Based on the dataset of course titles, descriptions, and tags, we train a word vector model to automatically extract synonym groups with high semantic similarity. Combined with manual review by domain experts, automatically generated synonym groups are supplemented or corrected.

[0009] Preferably, step S3 includes a retrieval recommendation mechanism based on user intent perception, and the mechanism includes: Based on the keywords, historical behaviors and contextual conditions entered by the user, the deep learning model BERT is used to predict the user's search intent. Prioritize the courses that best match the user's intent in the search results; When the user input conditions are ambiguous, multiple intent classification results are recommended.

[0010] Preferably, step S2 includes a retrieval efficiency optimization method based on a multi-layer index architecture, the method comprising: First-level index: construct a single-dimensional index of course data, including time index, difficulty index, and keyword index; Second-level index: builds a composite index based on the multi-dimensional conditions commonly used by users; Prioritize compound index matching during retrieval to reduce query calculations; Combined with high-frequency access statistics, frequently used data is loaded into the memory index first to optimize response speed.

[0011] Preferably, the multi-layer index architecture further includes a dynamic index loading mechanism, which includes: Regularly collect statistics on the condition combinations that users frequently use; Dynamically generate and load corresponding composite indexes for high-frequency condition combinations; A delayed loading strategy is used for indexes of low-frequency condition combinations to reduce system resource usage.

[0012] Preferably, the weight sorting algorithm in step S5 includes a result merging and deduplication mechanism, and the mechanism includes: In multi-layer index query results, merge the result sets returned by different index layers; De-duplicate the duplicate course data; Reorder the outputs based on the weight distribution results.

[0013] Preferably, the step S5 includes a highlight display function of the search results, and the highlight display function includes: Highlight the parts of the course data that match the keywords entered by the user, such as keywords in the course title and introduction; The matching score of each course data is displayed on the search result page, and a user feedback channel is provided to mark irrelevant results.

[0014] Preferably, step S3 further includes a result caching mechanism, which includes: Pre-cache search results for commonly used search condition combinations; A dynamic calculation strategy is used for the results of uncommon condition combinations, which are generated and cached only when requested by the user; Update cache data regularly to ensure the timeliness and accuracy of retrieval results.

[0015] The present invention provides a method for online retrieval of course data based on a database, which has the following beneficial effects: This database-based online course data retrieval method forms an overall technical system through multi-dimensional retrieval conditions, weighted sorting, fuzzy extension to intent perception and performance optimization; combines fuzzy rule extension, user intent perception and multi-layer index optimization to achieve intelligent and efficient retrieval; solves the problems of inaccurate retrieval and slow response speed in large-scale course data scenarios, and significantly improves user experience; solves the shortcomings of existing methods in dealing with complex needs, and significantly improves the accuracy and efficiency of retrieval. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 The present invention is a flowchart of a method for online retrieval of course data based on a database. DETAILED DESCRIPTION

[0017] The present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. The embodiments of the present invention are provided for the purpose of illustration and description, and are not intended to be exhaustive or to limit the present invention to the disclosed forms. Many modifications and variations will be apparent to those of ordinary skill in the art. The embodiments are selected and described in order to better illustrate the principles and practical applications of the present invention, and to enable those of ordinary skill in the art to understand the present invention and thereby design various embodiments with various modifications suitable for specific uses.

[0018] like Figure 1 As shown, the present invention provides a technical solution: a method for online retrieval of course data based on a database, comprising the following steps: S1: Construct a course database table, which includes the following fields: course name, course start time, difficulty level, and course keywords; S2: Define multi-dimensional search conditions, including time dimension, difficulty dimension and subject keyword dimension; S3: Provides a user front-end input interface for selecting or entering search conditions; S4: Based on the search conditions input by the user, dynamically generate database query statements, perform multi-dimensional screening, and obtain screening results; S5: Sort the screening results through a weight sorting algorithm, and give priority to displaying the courses that best match user needs.

[0019] The multi-dimensional search conditions in step S2 include: Time dimension: supports users to filter courses by specific time periods or cycles, such as weekends and evenings; Difficulty dimension: filter based on the difficulty level field in the course data, where the difficulty levels include beginner, intermediate, and advanced; Topic keyword dimension: supports users to enter keywords and filter based on full-text matching of course titles, introductions or tags.

[0020] The online course data retrieval method based on the database supports the retrieval of large-scale course data and adopts the following optimization strategies: Based on the full-text indexing technology Elasticsearch, fast keyword matching is achieved; Dynamically generate SQL query statements and improve database retrieval efficiency through joint indexes; For users' frequently used filtering combinations, some search results are cached in advance to obtain cache indexes, reducing the pressure of real-time calculations.

[0021] The step S3 includes introducing deep learning intention perception technology into the course retrieval scenario, and the process is as follows: User intent perception model: Based on the keywords, historical behaviors, and contextual conditions entered by the user, the deep learning model BERT is used to predict the user's search intent; For example, when a user searches for "Python", we can identify their preference for "elementary programming" or "data analysis" through their behavior history. Intelligent result recommendation: In the search results, the courses that best meet the user's intention are displayed first. For the conditions that the user does not specify, the classification results with different intentions are recommended to obtain the recommended results; for example: "Python beginner course"; "Python data analysis course"; Dynamic feedback optimization: Based on the user's click behavior on the recommended results, the prediction weight of the intent-aware model is adjusted to improve the accuracy of recommendations; During the retrieval process, users may fail to clearly express their complete intentions. For example, when searching for "Python", they do not indicate whether it is an elementary or advanced course, resulting in a search result range that is too large. In this regard, deep learning intent perception technology is introduced into the course retrieval scenario to solve the problem of low retrieval accuracy caused by ambiguous user input, while providing differentiated recommendations.

[0022] The weight sorting algorithm in step S5 includes the following steps: S51: The user sets a weight value for each dimension search condition, and the sum of the weight values ​​of each dimension is 1; S52: The calculation formula for the matching score of the search results is:

[0023] in, The weight set for the user; is the matching score of the corresponding dimension; Score the comprehensive matching degree of the search results, indicating the overall fit between the course and the user's needs, for sorting; The total number of dimensions of the screening conditions, including time dimension, difficulty dimension, and keyword dimension; The weight of the i-th dimension set by the user or adjusted dynamically; is the matching score of the i-th dimension, indicating the degree of matching between the course and the user's conditions in this dimension.

[0024] Specifically, when the user enters the following conditions: time dimension weight: 50%, matching score: 80%; difficulty dimension weight: 30%, matching score: 60%; keyword dimension weight: 20%, matching score: 90%; Using the matching score formula, the resulting score is 76%, and the system sorts the courses based on this score.

[0025] S53: Arrange the search results in descending order according to the matching scores.

[0026] The method further includes a step of dynamically adjusting the weight based on user behavior data, the step comprising: Behavioral data collection and analysis: Collect user behavior data after searching, including user click frequency, stay time, browsing, collection and registration operations; Build a user preference model based on the user's retrieval behavior data and extract the user's preference weights for different conditions; Weight dynamic optimization mechanism: When the user does not manually adjust the weight, the default weight is automatically adjusted according to the user's historical behavior model; for example: when the user clicks "Evening Course" multiple times, the system will increase the weight of the time dimension by default.

[0027] Weight optimization formula: ; in, It is the default dimension weight value, provided by the system when it is initialized; It is the user behavior feedback score, which indicates the user's preference for a certain dimension. The score is calculated based on the user's clicks, dwell time, collection or registration behavior. is the weight adjustment coefficient, which is used to adjust the influence of user behavior feedback on weight adjustment; its value is between 0 and 1 to avoid over- or under-adjustment; is the dimension weight value after dynamic adjustment, that is, the weight finally used for sorting.

[0028] Personalized weight learning: Introducing collaborative filtering algorithms from machine learning algorithms to learn the best strategy for weight adjustment from user data across the entire platform; For new users, initialize based on the user's basic information or the weight model of similar users; Among them, user behavior data includes clicks, browsing, and collections; basic information includes occupation and age; It should be further explained that, in the specific implementation process, users may have different priorities in terms of the dimensions of conditions they focus on in different search scenarios. For example, professional users may pay more attention to the time period, while students pay more attention to the difficulty of the course. The existing static weight distribution is difficult to meet the preferences of different user groups. In this regard, based on user behavior data: clicks, browsing, and collections; dynamically adjust the weight distribution rules, and achieve more accurate personalized search sorting through weight optimization driven by behavioral data, thereby improving user experience.

[0029] In step S3, an intelligent conditional expansion mechanism based on fuzzy rules is introduced, which mainly includes the following steps: S31: Conditional semantics extension: S311. Build keyword synonyms and related vocabulary, including expanding "data analysis" to "data visualization" and "data mining".

[0030] S312. When the search conditions are insufficient, the keyword field entered by the user is intelligently expanded to cover related courses.

[0031] S313. Calculate the semantic similarity of keywords based on the word vector model Word2Vec for dynamic expansion.

[0032] S32: Flexible adjustment of time interval: S321. When there are no classes in the user-specified time period, the time conditions are automatically relaxed; including: when the user selects Saturday, the system automatically expands it to "Saturday and Sunday".

[0033] S322. Assign different priorities to different time periods, including: a high-priority time period has a higher matching score.

[0034] S33: Fuzzy matching and rule fusion: S331. Add search conditions based on the results of fuzzy matching, and provide an optional "extended search mode" for users to switch.

[0035] It should be further explained that in actual searches, the search conditions entered by users may not be completely accurate or may not cover enough. For example, the keywords entered are too narrow: "data analysis" but not "data visualization", or the time selection is too strict: only Saturday is selected, which may lead to insufficient search results or even no results. The above technical solution introduces an intelligent condition expansion mechanism based on fuzzy rules, which can solve the problem of insufficient user condition input and significantly improve the coverage of search results while retaining accuracy. Compared with the existing simple condition matching mechanism, this expansion solution is more intelligent.

[0036] The keyword synonym library is generated by combining manual construction and automatic learning, including: Based on the dataset of course titles, descriptions, and tags, we train a word vector model to automatically extract synonym groups with high semantic similarity. Combined with manual review by domain experts, automatically generated synonym groups are supplemented or corrected.

[0037] Step S3 includes a retrieval recommendation mechanism based on user intent perception, which includes: Based on the keywords, historical behaviors and contextual conditions entered by the user, the deep learning model BERT is used to predict the user's search intent. Prioritize the courses that best match the user's intent in the search results; When the user input conditions are ambiguous, multiple intent classification results are recommended.

[0038] Step S2 includes a retrieval efficiency optimization method based on a multi-layer index architecture, the method comprising: S21: Multi-layer index design: The first layer: construct a single-dimensional index of the course field, including time index, difficulty index, and keyword index; The second layer: Based on the multi-dimensional condition combinations commonly used by users, a composite index is constructed, including the index of time and keywords; S22: Dynamic index loading: To meet real-time search requirements, dynamically load cache indexes of user-common condition combinations; When searching, the composite index of the second layer is matched first to reduce the amount of query calculation; S23: Hot data priority mechanism: Regularly collect statistics on frequently accessed course data and load them into the memory index first to improve retrieval speed; Use a delayed loading strategy for infrequently used data to optimize resource usage; S24: Result merging and sorting: Execute queries simultaneously in multi-layer indexes, and quickly merge and deduplicate result sets to reduce repeated calculations.

[0039] It should be further explained that, in the specific implementation process, in the large-scale course data scenario, the simple keyword index or time index in the single-layer index structure may face performance bottlenecks, especially when searching for multi-dimensional condition combinations, the response speed will drop significantly; in this regard, a retrieval efficiency optimization method based on a multi-layer index architecture is proposed, which solves the retrieval performance bottleneck problem under multi-dimensional condition combinations and significantly improves the system response speed.

[0040] The multi-layer index architecture further includes a dynamic index loading mechanism, which includes: Regularly collect statistics on the condition combinations that users frequently use; Dynamically generate and load corresponding composite indexes for high-frequency condition combinations; A delayed loading strategy is used for indexes of low-frequency condition combinations to reduce system resource usage.

[0041] The weight sorting algorithm in step S5 includes a result merging and deduplication mechanism, which includes: In multi-layer index query results, merge the result sets returned by different index layers; De-duplicate the duplicate course data; Reorder the outputs based on the weight distribution results.

[0042] Step S5 includes a highlighting function for the search results, and the highlighting function includes: Highlight the parts of the course data that match the keywords entered by the user, such as keywords in the course title and introduction; The matching score of each course data is displayed on the search result page, and a user feedback channel is provided to mark irrelevant results.

[0043] Step S3 also includes a result caching mechanism, which includes: Pre-cache search results for commonly used search condition combinations; A dynamic calculation strategy is used for the results of uncommon condition combinations, which are generated and cached only when requested by the user; Update cache data regularly to ensure the timeliness and accuracy of retrieval results.

[0044] It needs to be further explained that, in the specific implementation process, an overall technical system is formed through multi-dimensional retrieval conditions, weighted sorting, fuzzy expansion to intent perception and performance optimization; combined with fuzzy rule expansion, user intent perception and multi-layer index optimization, intelligent and efficient retrieval is achieved; the problems of inaccurate retrieval and slow response speed in large-scale course data scenarios are solved, and the user experience is significantly improved; comprehensive coverage from functional design to performance optimization provides a comprehensive course data retrieval solution for online education platforms.

[0045] It needs to be further explained that the database design: the database design includes the classification table and timetable in the course table and related auxiliary tables; the course table contains fields: course ID, name, time, difficulty, keywords, and ratings; through index optimization, it supports rapid retrieval of large-scale data.

[0046] Front-end interface interaction: Users select time, difficulty and input keywords through the interface; support free combination of screening conditions and setting priority weights for different conditions.

[0047] Weighted sorting implementation: In the search results, the matching score of each course is calculated based on the user weight and multi-dimensional matching score; the optimal result is output through the sorting algorithm; Performance optimization: In large-scale data scenarios, query speed is improved through caching and full-text indexing technology; at the same time, users' commonly used search combinations are recorded to optimize future queries.

[0048] The above solution can effectively improve the accuracy, efficiency and intelligence of the online course data retrieval system through in-depth exploration of four aspects: intelligent condition expansion, behavior data-driven weight adjustment, multi-layer index optimization and intent-aware recommendation.

[0049] It needs to be further explained that in the specific implementation process: For a working person who plans to learn data analysis skills on weekends, has limited time and wants the course difficulty to be elementary; After using the above search method, he entered into the platform: Time conditions: weekend; Difficulty level: beginner; Subject Keywords: Data analysis; The system is based on a comprehensive ranking of time priority and subject matching, where the weight of time priority is 50% and the weight of subject matching is 30%. The following search results are returned: 1. "Python Elementary Data Analysis", score: 95%; 2. "Basics of Data Analysis", score: 90%; 3. "Weekend Special Training: Introduction to Excel Data Analysis", score: 85%; This method not only improves the accuracy of user retrieval, but also significantly improves the user experience.

[0050] Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field and related fields without creative work should fall within the scope of protection of the present invention. The structures, devices and operating methods not specifically described and explained in the present invention are implemented according to the conventional means in the field unless otherwise specified and limited.

Claims

1. A method for online retrieval of course data based on a database, characterized in that: The following steps are involved: S1: Build a course database table. Includes the following fields: course name, course start time, difficulty level, course keywords; S2: Define multi-dimensional search conditions, including time dimension, difficulty dimension and subject keyword dimension; S3: Provides a user front-end input interface for selecting or entering search conditions; S4: Based on the search conditions input by the user, dynamically generate database query statements, perform multi-dimensional screening, and obtain screening results; S5: Sort the screening results through a weight sorting algorithm, and give priority to displaying the courses that best match user needs.

2. The online course data retrieval method based on a database according to claim 1, characterized in that: The method further includes a step of dynamically adjusting the weight based on user behavior data, the step comprising: Collect user behavior data after searching, including user click frequency, stay time, browsing, collection and registration operations; Build a user preference model based on user behavior data and extract the user's preference weights for different condition dimensions; When the user does not manually set the weights, the weights are dynamically adjusted according to the user preference model.

3. The online course data retrieval method based on a database according to claim 2 is characterized in that: In step S3, an intelligent condition extension mechanism based on fuzzy rules is introduced, and the intelligent condition extension mechanism includes the following steps: Use the keyword synonym library and word vector model Word2Vec to calculate the semantic similarity of the keywords entered by the user and dynamically expand the keyword field; When the time condition set by the user has no result, the time range is automatically relaxed to a nearby time period; Provide explicit identification of "extended mode" in search results, and allow users to choose to enable or disable extended mode.

4. The online course data retrieval method based on a database according to claim 3 is characterized by: The keyword synonym library is generated by combining manual construction and automatic learning, specifically including: Based on the dataset of course titles, descriptions, and tags, we train a word vector model to automatically extract synonym groups with high semantic similarity. Combined with manual review by domain experts, automatically generated synonym groups are supplemented or corrected.

5. The online course data retrieval method based on a database according to claim 4 is characterized in that: The step S3 includes a retrieval recommendation mechanism based on user intention perception, and the mechanism includes: Based on the keywords, historical behaviors and contextual conditions entered by the user, the deep learning model BERT is used to predict the user's search intent. Prioritize the courses that best match the user's intent in the search results; When the user input conditions are ambiguous, multiple intent classification results are recommended.

6. The online course data retrieval method based on a database according to claim 5 is characterized by: The step S2 includes a retrieval efficiency optimization method based on a multi-layer index architecture, the method comprising: First-level index: construct a single-dimensional index of course data, including time index, difficulty index, and keyword index; Second-level index: builds a composite index based on the multi-dimensional condition combinations commonly used by users; When searching, composite indexes are matched first; Combined with the high-frequency access statistics, frequently used data is loaded into the memory index first.

7. The online course data retrieval method based on a database according to claim 6 is characterized by: The multi-layer index architecture further includes a dynamic index loading mechanism, which includes: Regularly collect statistics on the condition combinations that users frequently use; Dynamically generate and load corresponding composite indexes for high-frequency condition combinations; A delayed loading strategy is used for indexes of low-frequency condition combinations to reduce system resource usage.

8. The online course data retrieval method based on a database according to claim 7 is characterized by: The weight sorting algorithm in step S5 includes a result merging and deduplication mechanism, and the mechanism includes: In multi-layer index query results, merge the result sets returned by different index layers; De-duplicate the duplicate course data; Reorder the outputs based on the weight distribution results.

9. The online course data retrieval method based on a database according to claim 8, characterized in that: The step S5 includes a highlighting function for displaying the search results, and the highlighting function includes: Highlight the parts of the course data that match the keywords entered by the user; The matching score of each course data is displayed on the search result page, and a user feedback channel is provided to mark irrelevant results.

10. The online course data retrieval method based on a database according to claim 9, characterized in that: The step S3 also includes a result caching mechanism, including: Pre-cache search results for commonly used search condition combinations; A dynamic calculation strategy is used for the results of uncommon condition combinations, which are generated and cached only when requested by the user; Update cached data periodically.

Images