A teaching content dynamic recommendation method based on interaction frequency

Abstract

This invention discloses a dynamic recommendation method for teaching content based on interaction frequency, relating to the field of educational informatization technology. This method collects interaction events generated by students and teachers during teaching sessions, and generates valid interaction events based on rules such as de-jittering, minimum dwell time, text length, repeated submission suppression, and abnormal high-frequency suppression. It uses a preset micro-window to statistically analyze individual interaction frequency, group interaction frequency, and class interaction frequency, and determines a target frequency band based on the teaching stage. When the class interaction frequency deviates from the target frequency band and meets the continuity condition, a trigger signal is generated. A strategy table is used to select the recommended action type, delivery scope, and time budget. Matching content is then selected from content items with increasing, stable, or decreasing frequency tags for delivery. After delivery, the class interaction frequency and low-frequency percentage are retested within the evaluation period to form an effect grading and audit record.

Description

Technical Field

[0001] This invention relates to the field of educational informatization technology, specifically to a method for dynamically recommending teaching content based on interaction frequency. Background Technology

[0002] While existing teaching platforms generally possess basic functions such as assignment posting, class attendance, in-class quizzes, discussion participation, resource access, and learning record statistics, their content recommendation systems largely rely on learning history, course preferences, knowledge mastery, question accuracy, response time, user profiles, or group similarity. The core objective of these systems is usually to recommend content that better suits learners' preferences or is more likely to improve grades, focusing on content matching rather than controlling the pace of the lesson. Because their decisions are based on multi-dimensional data, they often require complex data modeling and inference processes, making the rationale less intuitive. Teachers struggle to quickly understand the reasons for recommendations during class and find it difficult to immediately manage the results. Furthermore, while existing classroom interaction analysis systems track behavioral data such as the number of hands raised, questions answered, discussions, and resource clicks, these results are typically only used as input for post-class analysis reports, classroom activity scores, or learning progress warnings, failing to create a real-time closed loop of statistics, triggering, content delivery, and retesting. In other words, existing technologies can often only tell teachers whether there is too much or too little interaction, but they cannot immediately provide structured, actionable suggestions with a clear time budget for inserting teaching content when the interaction rhythm shifts, nor can they continue to retest its effectiveness after implementation. Summary of the Invention

[0003] The purpose of this invention is to provide a dynamic recommendation method for teaching content based on interaction frequency, in order to solve the problems in the existing technology of teaching content recommendation relying on complex models, lacking real-time rhythm control, unable to distinguish effective interactions, difficult to adapt to different teaching stages, unclear teacher control boundaries, and lack of audit closed loop.

[0004] To address the problems existing in the prior art, the present invention provides the following technical solution: a dynamic recommendation method for teaching content based on interaction frequency, the method being executed in a teaching session and including the following steps: S1, collecting interaction events generated between student and teacher terminals and writing them into an event log database, wherein each interaction event includes an event number, session number, user identifier, role identifier, event type, content number, occurrence time, terminal identifier, and additional data; S2, filtering the interaction events according to effective interaction judgment rules to generate effective interaction events; S3, statistically analyzing the effective interaction events according to a preset micro-window to generate individual interaction frequency, group interaction frequency, and class interaction frequency; S4, reading the current teaching stage and determining the corresponding target frequency band based on the current teaching stage, and generating a low-frequency proportion based on the individual interaction frequency and the lower bound of the target frequency band; S5, comparing the class interaction frequency with the... S6. When the frequency deviation state meets the continuous condition, a trigger signal is generated and written to the trigger log table; S7. Based on the current teaching stage, the frequency deviation state, the trigger signal, and the teaching situation snapshot, a strategy item is selected from the strategy table to obtain the recommended action type, the target object type, the target object range, and the time budget; S8. Content items that match the recommended action type and whose content effect tag direction corresponds to the frequency deviation state are selected from the content library to generate a recommendation instruction; S9. The recommendation instruction is delivered to the student end or presented to the teacher end according to the target range; the evaluation cycle is started after the teacher confirms the execution of the recommendation instruction or the system completes the delivery to the student end; S10. After the evaluation cycle ends, the class interaction frequency and low frequency ratio are retested, an effect rating is generated, and written to the statistical sedimentation table and the audit log table.

[0005] Preferably, the effective interaction determination rules include: when the same user ID triggers the same event type for the same content number and the time interval between the two occurrences is less than 2 seconds, only one interaction event is retained; after a resource opening event occurs, if the corresponding user leaves the content number within 2 seconds and does not generate a subsequent interaction event, the resource opening event is not counted as a valid interaction event; when the text length of a text answer submission event or discussion post is less than 2 Chinese characters or less than 4 characters, the text answer submission event or discussion post is not counted as a valid interaction event; when the same user ID generates multiple question submission events for the same question within 60 seconds, only the first question submission event is counted as a valid interaction event; when the same user ID generates 10 or more interaction events within 5 seconds, and the total proportion of resource opening events and prompt request events is not less than 70%, all interaction events within the 5 seconds are marked as abnormal events and removed from the valid interaction events.

[0006] Preferably, the micro-window is 120 seconds, the refresh cycle of the class interaction frequency is 10 seconds, and the trend statistics window is 600 seconds; the class interaction frequency is the median of the individual interaction frequency of all students; the low-frequency proportion is the proportion of students whose individual interaction frequency is lower than the lower bound of the target frequency band to the number of effective online students; the target frequency band includes: 0.8 times / minute to 2.0 times / minute for the explanation stage, 2.0 times / minute to 4.0 times / minute for the practice stage, 3.0 times / minute to 6.0 times / minute for the discussion stage, and 0.2 times / minute to 1.0 times / minute for the silent task stage.

[0007] Preferably, the triggering signals include a first-level low-frequency triggering signal, a first-level high-frequency triggering signal, a second-level stage switching suggestion triggering signal, and a third-level suppression triggering signal, wherein: a first-level low-frequency triggering signal is generated when the class interaction frequency is below the lower limit of the target frequency band for two consecutive micro-windows; a first-level high-frequency triggering signal is generated when the class interaction frequency is above the upper limit of the target frequency band for two consecutive micro-windows; a second-level stage switching suggestion triggering signal is generated when a first-level low-frequency triggering signal or a first-level high-frequency triggering signal has been generated, and the class interaction frequency has not returned to the target frequency band within three consecutive micro-windows, and the teacher has not enabled stage locking; a third-level suppression triggering signal is generated and automatic delivery is paused for 300 seconds when the number of students hitting the abnormal event accounts for no less than 20% of the number of valid online students, or when the event access delay exceeds 5 seconds three times within 60 seconds.

[0008] Preferably, each strategy entry in the strategy table includes a strategy number, teaching stage, frequency deviation status, trigger signal type, context gating condition, recommended action type, target audience type, target audience range, time budget, cooldown time, priority, and explanation template; the target audience type includes teacher-side suggestion type and student-side execution type; the target audience range includes the whole class, group, and individual; each content entry in the content library includes a content number, content type, knowledge tag, prerequisite tag, estimated time, content effect tag, effect intensity level, applicable stage, version number, and release status; the content effect tag includes frequency increase tag, frequency stability tag, and frequency decrease tag; when the frequency deviation status is below the target frequency band, content entries with the frequency increase tag are selected; when the frequency deviation status is within the target frequency band, content entries with the frequency stability tag are selected; when the frequency deviation status is above the target frequency band, content entries with the frequency decrease tag are selected; the content entries also meet the conditions that the estimated time is not greater than the time budget, the applicable stage is consistent with the current teaching stage, the prerequisite tag is satisfied, and it has not been deployed in the last 20 minutes.

[0009] Preferably, the evaluation period is 180 seconds, and the effect classification includes effective, moderate, and ineffective, wherein: when the evaluation period ends, if the class interaction frequency increases by at least 1.0 times / minute compared to before the trigger, or the proportion of low frequency decreases by at least 15 percentage points compared to before the trigger, the effect classification is effective; when the evaluation period ends, if the class interaction frequency does not increase by 1.0 times / minute, but the class interaction frequency has returned to the target frequency band, the effect classification is moderate; when the evaluation period ends, if the class interaction frequency is still below the lower limit of the target frequency band, or still above the upper limit of the target frequency band, or if a level 3 suppression trigger signal is generated during the evaluation period, the effect classification is ineffective.

[0010] A dynamic recommendation system for teaching content based on interaction frequency includes: an event access module for collecting interaction events generated between students and teachers and writing them into an event log database; an effective interaction judgment module for generating effective interaction events according to effective interaction judgment rules; an interaction frequency statistics module for calculating individual interaction frequency, group interaction frequency, class interaction frequency, and low-frequency percentage; a teaching context management module for generating a snapshot of the teaching context including the current teaching stage, stage locking status, silent task status, and remaining time; a trigger gating module for generating trigger signals based on the comparison results between class interaction frequency and target frequency band; a strategy table decision module for selecting strategy items based on the current teaching stage, frequency deviation status, trigger signals, and teaching context snapshot; a content selection module for selecting content items and generating recommendation instructions based on the recommended action type and content effect tags; a delivery and presentation module for delivering recommendation instructions to students or presenting them to teachers; a feedback and evaluation module for retesting class interaction frequency and low-frequency percentage after the evaluation period and generating an effect rating; and an audit and evidence storage module for storing event logs, trigger logs, recommendation logs, evaluation results, and operation traces.

[0011] Compared with the prior art, the present invention has the following beneficial effects: (1) The present invention does not rely on complex mechanisms such as knowledge mastery prediction, user profile, group similarity, and implicit interest mining, but only uses the single main control variable of interaction frequency to regulate the teaching rhythm. Therefore, the present invention has fewer input dimensions, shorter interpretation paths, and transparent operating logic, and teachers can quickly understand why the system suggests inserting a certain teaching content, thereby significantly improving the acceptability of the system in real classrooms; (2) The present invention first converts the original interactive events into effective interactive events, and then generates the interaction frequency, suppressing repeated clicks, short stays, empty text, and high-frequency refresh event noise from the source. This makes the interaction frequency on which the system is based more realistically reflect the classroom participation status, avoiding the erroneous judgment in the prior art that the system thinks it is very active, but in fact it is just repeatedly opening the page; (3) The present invention sets different target frequency bands for the explanation stage, practice stage, discussion stage, and silent task stage, so that the system has the ability to understand the classroom status according to the stage. This system will not misjudge silent thinking as low participation, nor will it misjudge normal high-frequency speaking during the discussion phase as an out-of-control state, significantly improving the accuracy of rhythm judgment; (4) This invention divides the triggering process into three levels: first-level fine-tuning, second-level stage suggestions, and third-level anomaly suppression, so that the system will not frequently interrupt teaching due to short-term fluctuations, nor will it continue to push content incorrectly under abnormal data conditions. Compared with the two-stage solutions in the existing technology that push content as soon as it deviates or only analyzes without intervention, it can achieve a better balance between automation and classroom stability; (5) This invention assigns three types of content effect labels to content items: frequency increase, frequency stability, and frequency decrease, so that content selection is no longer just content matching, but rhythm regulation execution. For the same teaching content item, you not only need to know what it talks about, but also which direction it will push the classroom rhythm. This gives the content library rhythm function attributes; (6) This invention sets a 180-second evaluation period after the content is delivered, retests the frequency of class interaction and the proportion of low frequency before and after delivery, and forms three levels of results: effective, average, and ineffective. This not only verifies whether the recommendation truly improves the classroom rhythm, but also provides a direct basis for subsequent content library annotation and strategy table optimization, overcoming the problem of broken links that end after the recommendation in existing technologies; (7) This invention, through stage locking, silent task switching, automatic delivery pause, manual delivery, replacement, and withdrawal mechanisms, always places the teacher in the ultimate leading position. The system is only responsible for providing structured suggestions and automated support, without depriving the teacher of control over the classroom rhythm, thus better meeting the usage needs in real teaching scenarios; (8) This invention records all interactive events, effective interaction judgments, frequency statistics, trigger signals, strategy items, content items, delivery results, and evaluation results, so the cause, execution path, and effect of any recommendation can be traced back. It not only has teaching application value, but also engineering verifiability and subsequent maintenance convenience. Attached Figure Description

[0012] Figure 1 This is a schematic diagram of the method flow of the present invention. Detailed Implementation

[0013] The technical solution of the present invention will be further described below with reference to the accompanying drawings and specific embodiments.

[0014] This invention proposes a dynamic recommendation method for teaching content based on interaction frequency, which runs in the teaching session and unfolds around a closed-loop chain: interaction event collection, effective interaction judgment, interaction frequency statistics, teaching context identification, deviation triggering, strategy table decision, content item selection, content delivery, and effect retesting.

[0015] A teaching session is a complete classroom session or a complete online learning task; interactive events are discrete operation records generated by students or teachers during the teaching session; effective interactive events are interactive events that are retained after being filtered by rules and can truly reflect participation behavior; interaction frequency is the result of counting and converting effective interactive events within a preset sliding window; target frequency band is a preset reasonable interaction frequency range for the corresponding teaching stage; frequency deviation status is the comparison result of interaction frequency relative to the target frequency band; trigger signal is a discrete control signal generated when the frequency deviation status continuously meets the conditions; content effect label is the expected direction of influence of content items on the interaction rhythm; evaluation period is a fixed time period used to retest the effect after content is released.

[0016] Specifically, the present invention includes the following steps: Standardized collection of interactive events: During teaching sessions, student and teacher terminals continuously generate activity records, which are uniformly written into the event log database.

[0017] Each interactive event includes the following fields: event number; session number; user ID; role ID; event type; content number; occurrence time; terminal ID; additional data.

[0018] The event type is fixedly defined as one of the following sets: Question submission event; vote submission event; text answer submission event; hand raising event; discussion speaking event; resource opening event; annotation addition event; prompt request event; teacher interaction posting event; teacher roll call posting event.

[0019] Rule-based determination of valid interactive events: In order to prevent invalid clicks, repeated actions, and event-brushing behaviors from interfering with the judgment of classroom rhythm, this invention does not directly use the original interactive events, but first filters them through valid interaction determination rules.

[0020] The fixed judgment rules include: De-jittering rule: When the same user ID triggers the same event type for the same content number, and the time interval between the two occurrences is less than 2 seconds, only 1 interaction event is retained as a valid interaction event.

[0021] Minimum dwell time rule: If a user leaves the corresponding content within 2 seconds after a resource open event occurs and no subsequent events are generated, the resource open event will not be counted as a valid interaction event.

[0022] Text length rule: Text content length of text response submission events and discussion posting events that is less than 2 Chinese characters or less than 4 characters will not be counted as valid interaction events.

[0023] Duplicate submission suppression rule: When the same user submits multiple questions for the same question within 60 seconds, only the first submission is counted as a valid interaction event.

[0024] Abnormal high-frequency suppression rule: If the same user generates 10 or more candidate interaction events within 5 seconds, and the total proportion of resource opening events and prompt request events is not less than 70%, all candidate interaction events within those 5 seconds will be marked as abnormal events and will not be counted as valid interaction events.

[0025] By using the above rules, the original number of interactions can be transformed into effective interactions that have educational significance.

[0026] The hierarchical statistical mechanism for interaction frequency sets up three levels of frequency statistics objects: individual interaction frequency; group interaction frequency; and class interaction frequency. Among them: individual interaction frequency is used for fine-grained identification of individual students; group interaction frequency is used for group collaboration scenarios; and class interaction frequency is used for rhythm scheduling of the entire class.

[0027] Three time scales are further defined: Micro-window: 120 seconds; Refresh cycle: 10 seconds; Trend window: 600 seconds. The system refreshes the statistical results of valid interactive events in the current micro-window every 10 seconds; the system refreshes the statistical results of trends in the trend window every 60 seconds.

[0028] At the class level, the class interaction frequency is calculated using the median of individual student interaction frequencies to avoid individual outliers from affecting the overall assessment. Additionally, the system generates a low-frequency percentage, defined as the proportion of students whose frequency is below the lower bound of the current stage's target frequency to the total number of active online students.

[0029] The teaching phased target frequency band divides the teaching conversation into four basic phases: explanation phase; practice phase; discussion phase; and silent task phase.

[0030] The target frequency bands for each stage are fixedly set as follows: Explanation stage: 0.8 to 2.0 times / minute; Practice stage: 2.0 to 4.0 times / minute; Discussion stage: 3.0 to 6.0 times / minute; Silent task stage: 0.2 to 1.0 times / minute.

[0031] Through the above design, the system no longer uniformly identifies all low-interaction states as abnormal, but instead uses the teaching stage as the basis for judgment.

[0032] The tiered trigger gating mechanism does not adopt a crude approach of pushing out controls as soon as they deviate, but instead uses a three-tiered gating structure: Level 1 Trigger Signal: Used for short-term fine-tuning, it is divided into Level 1 low-frequency trigger signal and Level 1 high-frequency trigger signal. Generation conditions: When the class interaction frequency is below the lower bound of the target frequency band for two consecutive micro-windows, a Level 1 low-frequency trigger signal is generated; when the class interaction frequency is above the upper bound of the target frequency band for two consecutive micro-windows, a Level 1 high-frequency trigger signal is generated.

[0033] Second-level stage switching suggestion trigger signal: Used to prompt teachers to adjust teaching organization. Generation conditions: After the first-level trigger signal has been generated; in the subsequent three consecutive micro-windows, the class interaction frequency has not returned to the target frequency band; and the teacher has not enabled stage locking. When the above conditions are met, the second-level stage switching suggestion trigger signal is generated.

[0034] Level 3 Suppression Trigger Signal: Used to prevent abnormal data from causing system malfunctions. Generation conditions: The number of students hitting the high-frequency suppression rule accounts for at least 20% of the class's valid online students; or the event access delay exceeds 5 seconds three times within 60 seconds. When the above conditions are met, a Level 3 suppression trigger signal is generated, and automatic content delivery is paused for 300 seconds.

[0035] Strategy Table Decision-Making Mechanism: The dynamic recommendation in this invention is not predictive recommendation, but rather deterministic recommendation driven by a strategy table. Each strategy entry in the strategy table contains the following fields: strategy number; teaching stage; frequency deviation state; trigger signal type; context gating condition; recommended action type; delivery scope; time budget; cooldown time; priority; and explanation template. The recommended action type is fixedly defined as one of the following sets: voting task; in-class quiz task; example explanation card; discussion task; summary card; and silent task.

[0036] The system matches strategy entries based on teaching stage, frequency deviation status, trigger signal type, and context gating conditions, and selects the strategy entry with the highest priority for execution.

[0037] The content item effect tagging mechanism: Content items are not merely viewed as knowledge carriers, but rather each item is further imbued with an interactive rhythm effect. Each content item must contain at least the following fields: Content ID; Content Type; Knowledge Tag; Pre-existing Tag; Estimated Duration; Content Effect Tag; Effect Intensity Level; Applicable Stage; Distribution Scope; Version Number; Release Status. The content effect tag is fixed as one of the following three types: Frequency Increase Tag; Frequency Stabilization Tag; Frequency Decrease Tag. The effect intensity level is fixed as: Level 1; Level 2; Level 3.

[0038] When the class interaction frequency is below the target frequency band, the system selects only content items with the "increased frequency" tag; when the class interaction frequency is within the target frequency band, the system selects only content items with the "stable frequency" tag; when the class interaction frequency is above the target frequency band, the system selects only content items with the "decreased frequency" tag. Additionally, the selected content items must meet the following constraints: the applicable stage must be consistent with the current teaching stage; the estimated time used must not exceed the time budget of the strategy item; the prerequisite tags must be met; and the content must not have been re-delivered within the last 20 minutes.

[0039] Teacher-led control retention mechanism: Control items set on the teacher's end: stage selector; stage lock switch; silent task switch; automatic delivery pause button; manual delivery button; replace content button; withdraw content button.

[0040] When a teacher enables stage locking, the system will not automatically deliver content to students, but will only display suggestions to the teacher. When a teacher enables a silent task, the system will switch to the silent task stage target frequency band and stop triggering the first-level frequency increase. When a teacher clicks the pause button, the system will stop automatically delivering content for 300 seconds.

[0041] Post-launch evaluation loop: After content launch, the system immediately enters a 180-second evaluation period. At the end of the evaluation period, the system retests the class interaction frequency and low-frequency percentage, and classifies the effect according to the following rules: Effective: Class interaction frequency increases by at least 1.0 times / minute compared to before triggering, or the low-frequency percentage decreases by at least 15 percentage points compared to before triggering; Average: Class interaction frequency does not increase by 1.0 times / minute, but has returned to the target frequency band; Ineffective: Class interaction frequency is still below the lower limit of the target frequency band, or still above the upper limit of the target frequency band, or a level 3 suppression trigger signal is generated.

[0042] All recommendation results are written to the statistical sedimentation table and audit log table for subsequent manual calibration of content effect labels and strategy items.

[0043] Dynamic recommendation in basic classroom scenarios: Teaching session initialization: After the teacher starts a class, the system generates a unique session number and creates an event log table, frequency statistics table, trigger log table, recommendation log table, and evaluation result table for this teaching session. The teacher selects the current teaching stage as the practice stage on the teacher side, and the system reads the corresponding target frequency band for this stage as 2.0 times per minute to 4.0 times per minute.

[0044] Interaction event generation and reporting: During the practice process, the student side generates question submission events, hint request events, and text answer submission events. The student side batches and reports the local event cache to the event access gateway every 10 seconds. The event access gateway de-duplicates the event numbers and writes them into the event log library according to the occurrence time.

[0045] Effective interaction determination: The system performs the following determinations on the original interaction events in the event log library: If a student clicks on the same resource page three times continuously, and the interval between each click is less than 2 seconds, only the first resource opening event is retained; if a student submits text with a length less than 2 Chinese characters, such as "um" or "good", it is not counted as an effective interaction event; if a student submits answers to the same question three times continuously within 30 seconds, only the first question submission event is retained; if a student generates 12 hint request events continuously within 5 seconds, all related events within these 5 seconds are excluded, and an exception prompt is generated. After processing, the system only retains real and effective interaction events.

[0046] Frequency statistics: The system takes 120 seconds as the micro-window and 10 seconds as the refresh period to count the individual interaction frequencies of each student. Subsequently, the median of the individual interaction frequencies of all students is used as the class interaction frequency. If there are 40 effective online students in a class, and the individual interaction frequencies of 26 students are lower than 2.0 times per minute, then the low-frequency ratio is 65%.

[0047] Primary trigger generation: If the class interaction frequency is 1.1 times per minute and 1.0 times per minute in two consecutive micro-windows, both of which are lower than the lower bound of the target frequency band of 2.0 times per minute in the practice stage, the system generates a primary low-frequency trigger signal.

[0048] Strategy table matching: The system matches strategy entries according to the following conditions: Teaching stage: practice stage; Frequency deviation status: lower than the target frequency band; Trigger signal: primary low-frequency trigger signal; Context gating condition: remaining time is greater than 8 minutes; Stage lock closed; Silent task closed. The entry matched by the strategy table is: Recommended action type: voting task; Delivery object type: student-side execution type; Delivery object range: the whole class; Time budget: 60 seconds; Cooling time: 600 seconds; Priority: 1.

[0049] Content item selection: The system searches the content library for content items that meet the following conditions: the content type is a voting task; the content effect tag is an upscaling tag; the applicable stage includes the practice stage; the estimated time is no more than 60 seconds; it has not been deployed in the last 20 minutes; and the prerequisite tags are met. The system ultimately selects the voting task with content number C102 and generates a recommendation number R10201.

[0050] Content delivery: The teacher's recommendation reason is: During the practice phase, the class interaction frequency was below the target frequency (lower bound) for 4 consecutive minutes, with low frequency accounting for 65%. It is recommended to deliver a 60-second frequency-boosting voting task. Since the teacher did not enable phase locking, the system simultaneously delivered the C102 voting task to all students' devices.

[0051] Effectiveness retest: After the campaign ended, the system entered a 180-second evaluation period. At the end of the evaluation period, the frequency of class interactions increased from 1.0 times / minute to 2.3 times / minute, and the percentage of low-frequency interactions decreased from 65% to 42%. The system marked this recommendation as effective.

[0052] Misjudgment suppression in silent reading scenarios: Teacher switching phase: The teacher arranges 3 minutes of silent reading in class and switches the current phase to the silent task phase on the teacher's device.

[0053] Target frequency band change: The system automatically switches the target frequency band from 0.2 times / minute to 1.0 times / minute. At the same time, the system disables the first-level low-frequency trigger signal generation logic, retaining only the third-level suppression trigger signal monitoring.

[0054] Rhythm recognition results: Although the frequency of classroom interaction decreased to 0.4 times / minute, it remained within the target frequency band for the silent task phase, so the system did not intervene to increase the frequency. This avoids misjudging normal quiet reading as dull classroom behavior.

[0055] Security Suppression in Abnormal Event Scenarios: An abnormal event occurred: During a teaching session, 8 students repeatedly triggered resource open events and prompt request events within 5 seconds due to network jitter or accidental operation. These resource open events and prompt request events accounted for 80% of the total number of events for these students.

[0056] Abnormal high-frequency suppression rule hit: The system will classify all candidate interaction events within the above 5 seconds as abnormal events and will not count them as valid interaction events. At the same time, since the number of abnormal students accounts for more than 20% of the number of valid online students, the system will generate a level 3 suppression trigger signal.

[0057] Automatic delivery pause: After the level 3 suppression trigger signal is generated, the system pauses automatic delivery for 300 seconds, only notifying the teacher that an abnormal high-frequency event has been detected and automatic delivery suppression has been entered. This avoids abnormal data driving incorrect teaching content recommendations.

[0058] Frequency reduction adjustment during the discussion phase: Discussion phase initiated: Teachers switch the teaching phase to the discussion phase, and the target frequency band is automatically set to 3.0 times / minute to 6.0 times / minute.

[0059] High-frequency state appears: In two consecutive micro-windows, the class interaction frequency reached 6.8 times / minute and 7.1 times / minute respectively, both exceeding the upper limit of the target frequency band. The system generated a first-level high-frequency trigger signal.

[0060] Strategy Item Matching: The strategy items matched by the system are: Recommended Action Type: Summary Card; Target Audience Type: Teacher-Suggested Type; Target Audience Scope: Whole Class; Time Budget: 120 seconds; Content Effect Tag: Frequency Reduction Tag.

[0061] Teacher's suggestion: The teacher received a suggestion: The interaction frequency in the current discussion phase has continuously exceeded the upper limit. It is recommended to insert a 120-second summary card to conclude the discussion and return to a unified pace. After the teacher confirms, the system distributes the summary card to the whole class.

[0062] Retest results: After 180 seconds, the frequency of class interaction dropped back to 5.1 times / minute, returning to the target frequency band for the discussion phase. The system recorded the effect of this campaign as valid.

Claims

1. A method for dynamically recommending teaching content based on interaction frequency, characterized in that, The method is performed during a teaching session and includes the following steps: S1. Collect the interaction events generated between the student terminal and the teacher terminal, and write them into the event log library. Each interaction event includes an event number, session number, user ID, role ID, event type, content number, occurrence time, terminal ID, and additional data. S2. Filter the interactive events according to the valid interaction judgment rules to generate valid interactive events; S3. Statistically analyze the effective interactive events according to the preset micro-window to generate individual interaction frequency, group interaction frequency, and class interaction frequency; S4. Read the current teaching stage, determine the corresponding target frequency band based on the current teaching stage, and generate the low frequency ratio based on the individual interaction frequency and the lower bound of the target frequency band; S5. Compare the class interaction frequency with the target frequency band to generate a frequency deviation status; S6. When the frequency deviation state meets the continuous condition, a trigger signal is generated and written to the trigger log table; S7. Based on the current teaching stage, the frequency deviation state, the trigger signal, and the teaching situation snapshot, select strategy items from the strategy table to obtain the recommended action type, target type, target range, and time budget. S8. Filter content items from the content library that match the recommended action type and whose content effect tag direction corresponds to the frequency deviation state, and generate a recommendation instruction; S9. Deliver the recommendation instruction to the student terminal or present the recommendation instruction to the teacher terminal according to the delivery range; start the evaluation cycle after the teacher confirms the execution of the recommendation instruction or the system completes the delivery to the student terminal; S10. After the evaluation period ends, retest the frequency of class interactions and the proportion of low-frequency interactions, generate an effectiveness classification, and write it into the statistical record table and audit log table.

2. The method for dynamically recommending teaching content based on interaction frequency according to claim 1, characterized in that, The rules for determining valid interaction include: If the same user ID triggers the same event type for the same content number and the time interval between the two occurrences is less than 2 seconds, only one interaction event will be retained. If, after a resource open event occurs, the corresponding user leaves the content number within 2 seconds and no subsequent interaction event occurs, the resource open event will not be counted as a valid interaction event. If the text length of a text answer submission event or discussion post event is less than 2 Chinese characters or less than 4 characters, the text answer submission event or discussion post event will not be counted as a valid interaction event. When the same user ID submits multiple questions for the same question within 60 seconds, only the first question submission event is counted as a valid interaction event. If the same user ID generates 10 or more interactive events within 5 seconds, and the total proportion of resource open events and prompt request events is not less than 70%, all interactive events within the 5 seconds will be marked as abnormal events and removed from the valid interactive events.

3. The method for dynamically recommending teaching content based on interaction frequency according to claim 1, characterized in that: The micro-window lasts for 120 seconds, the refresh cycle for the class interaction frequency is 10 seconds, and the trend statistics window lasts for 600 seconds. The class interaction frequency is the median of the individual interaction frequency of all students; The low-frequency percentage is the proportion of students whose individual interaction frequency is lower than the lower bound of the target frequency to the total number of valid online students. The target frequency bands include: 0.8 to 2.0 times per minute for the explanation phase, 2.0 to 4.0 times per minute for the practice phase, 3.0 to 6.0 times per minute for the discussion phase, and 0.2 to 1.0 times per minute for the silent task phase.

4. The method for dynamically recommending teaching content based on interaction frequency according to claim 1, characterized in that, The trigger signals include a first-level low-frequency trigger signal, a first-level high-frequency trigger signal, a second-level stage switching suggestion trigger signal, and a third-level suppression trigger signal, wherein: When the class interaction frequency is below the lower bound of the target frequency band for two consecutive micro-windows, a level 1 low-frequency trigger signal is generated. When the frequency of class interaction exceeds the upper limit of the target frequency band for two consecutive micro-windows, a first-level high-frequency trigger signal is generated. When a Level 1 low-frequency trigger signal or a Level 1 high-frequency trigger signal has been generated, and the class interaction frequency has not returned to the target frequency band within the next three consecutive micro-windows, and the teacher has not enabled stage locking, a Level 2 stage switching suggestion trigger signal is generated. When the number of students who trigger abnormal events accounts for no less than 20% of the number of valid online students, or when the event access delay exceeds 5 seconds three times within 60 seconds, a level 3 suppression trigger signal is generated, and automatic delivery is suspended for 300 seconds.

5. The method for dynamically recommending teaching content based on interaction frequency according to claim 1, characterized in that, Each strategy entry in the strategy table includes a strategy number, teaching stage, frequency deviation status, trigger signal type, context gating condition, recommended action type, target type, target range, time budget, cooldown time, priority, and explanation template. The types of targets for delivery include teacher-suggested suggestions and student-implemented actions; the scope of targets includes the whole class, small groups, and individuals. Each content entry in the content library includes a content number, content type, knowledge tag, prerequisite tag, estimated time, content effect tag, effect intensity level, applicable stage, version number, and release status. The content effect tags include upsampling tags, stable frequencies tags, and downsampling tags; When the frequency deviation is below the target frequency band, filter the entries with the up-frequency tag. When the frequency deviation status is within the target frequency band, filter the stable frequency tag content items; When the frequency deviation is higher than the target frequency band, filter out the entries with the down-frequency tag; The content items also meet the following conditions: the estimated time taken is no more than the time budget, the applicable stage is consistent with the current teaching stage, the pre-requirement tags are met, and the content has not been displayed in the last 20 minutes.

6. The method for dynamically recommending teaching content based on interaction frequency according to claim 1, characterized in that, The evaluation period is 180 seconds, and the effect grading includes effective, average, and ineffective, wherein: When the evaluation period ends, if the frequency of class interaction increases by at least 1.0 times / minute compared to before the trigger, or the proportion of low frequency decreases by at least 15 percentage points compared to before the trigger, the effect is rated as effective. If, at the end of the evaluation period, the frequency of class interaction has not increased by 1.0 times / minute, but has returned to the target frequency band, the effect is classified as average. If, at the end of the evaluation period, the frequency of class interaction is still below the lower bound of the target frequency band, or still above the upper bound of the target frequency band, or if a level 3 inhibition trigger signal is generated during the evaluation period, the effect is classified as invalid.

7. A dynamic recommendation system for teaching content based on interaction frequency, characterized in that, include: The event access module is used to collect interactive events generated between students and teachers and write them to the event log library; The effective interaction determination module is used to generate effective interaction events based on the effective interaction determination rules. The interaction frequency statistics module is used to count individual interaction frequency, group interaction frequency, class interaction frequency, and the proportion of low-frequency interactions. The teaching context management module is used to generate a snapshot of the teaching context, which includes the current teaching stage, stage locking status, silent task status, and remaining time. The trigger gating module is used to generate trigger signals based on the comparison results between the class interaction frequency and the target frequency band; The strategy table decision module is used to select strategy items based on the current teaching stage, frequency deviation status, trigger signals, and snapshots of the teaching context. The content selection module is used to select content items and generate recommendation instructions based on the recommended action type and content effect tags. The delivery and presentation module is used to deliver recommendation instructions to students or present them to teachers. The feedback and evaluation module is used to retest the frequency of class interactions and the proportion of low-frequency interactions after the evaluation period ends and generate an effectiveness rating. The audit and evidence storage module is used to save event logs, trigger logs, recommendation logs, evaluation results, and operation traces.

Images