Mobile English learning system based on interaction

Abstract

The invention relates to the technical field of online learning, in particular to an interaction-based mobile English learning system, which comprises a grammar error interaction recognition module, an error item pushing frequency regulation and control module, a learning load dynamic monitoring module, a learning difficulty dynamic adjustment module and an English skill interaction aggregation module. According to the method, the error event density change in the English learning interaction information of the user is dynamically monitored, the error event time sequence is established by using the error labeling field, the error type classification and the timestamp, and the learning content pushing frequency adjustment object is accurately determined according to the error density change amplitude, so that efficient error item accurate intensified training is realized; meanwhile, according to the synchronous change amplitude of the task response time increase rate and the answer accuracy change rate, the change condition of the learning load of the user is recognized in real time, the difficulty level of the learning content is dynamically and automatically adapted, dynamic regulation and control between the learning difficulty and the learning state of the user are formed, and the learning acceptability and effect of the user are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of online learning, and particularly to an interactive mobile English learning system. Background Art

[0002] The technical field of online learning involves methods and systems for educational resource management, knowledge transfer, learning process interaction control, and learning effect evaluation in an Internet environment. This field includes multiple aspects such as learning content management, personalized learning path recommendation, online testing and feedback mechanisms, learning behavior data analysis, and intelligent assisted teaching. Online learning technology usually relies on specific technical means such as cloud computing, artificial intelligence, and big data analysis, and adopts technical means such as video teaching, interactive question banks, real-time answering, and speech recognition to achieve dynamic push of teaching content, personalized regulation of learning paths, and quantitative evaluation of learning effects.

[0003] Among them, the mobile English learning system is an English learning system based on mobile terminal devices, aiming to meet the needs of learners to flexibly and efficiently obtain language learning resources in different scenarios. Its purpose is to provide diverse learning methods, including text learning, voice interaction, listening training, scenario simulation exercises, etc., enabling users to continuously accumulate vocabulary, train grammar, and practice oral expression during fragmented time, thereby improving the comprehensive English application ability.

[0004] When traditional learning systems process learners' error feedback and interaction data, they only perform static error type identification and unified content push, ignoring the differences in the density change and load fluctuation of different error types during the user's learning process, resulting in insufficient adaptation of learning content to the user's real-time state, poor matching between learning load and learning difficulty, and prone to situations where the content is too difficult or too easy, leading to a decrease in learning enthusiasm and difficulty in significantly improving learning effects. Traditional systems mainly evaluate based on overall learning performance or the scores of a single skill, and fail to fully consider the differences in the interactive contribution degrees between skills, making it difficult to objectively quantify and comprehensively evaluate the actual mastery of learners' English skills, thus weakening the accuracy and pertinence of personalized learning path recommendation and resulting in limited user learning efficiency and skill improvement. Summary of the Invention

[0005] The purpose of the present invention is to solve the deficiencies existing in the prior art, and propose an interactive mobile English learning system.

[0006] To achieve the above purpose, the present invention adopts the following technical solutions: An interactive mobile English learning system, the system includes: The grammar error interaction recognition module obtains the English learning interaction information of the user, compares the cumulative value differences of adjacent two time windows based on the cumulative value of the error occurrence frequency within the preset time window, screens the set of error types whose density change amplitude exceeds the set error density change threshold, and establishes a grammar interaction error feature list; The wrong item push frequency regulation module calls the grammar interaction error feature list, and adjusts the push frequency of the English learning content corresponding to the error according to the error occurrence frequency and the original set push reference frequency, and generates an interactive wrong item push frequency adjustment set; The learning load dynamic monitoring module calls the interactive wrong item push frequency adjustment set, obtains the interactive response time and the answer correct rate identifier recorded during the interaction of the English learning content, and screens the set of tasks whose synchronous change amplitude exceeds the set synchronous change threshold based on the response time growth rate and the correct rate change rate of three consecutive tasks, and generates an interactive learning load change set; The learning difficulty dynamic adjustment module calls the interactive learning load change set, and adjusts the difficulty configuration of the pushed English learning content according to the proportion of the number of learning load abnormal tasks, and generates the dynamically adjusted content difficulty configuration information.

[0007] The improvements of the present invention are that the grammar interaction error feature list includes the error type density distribution, the error event time sequence trajectory and the error high-frequency word category, the interactive wrong item push frequency adjustment set is specifically a frequency supplement task set, a frequency reduction task set and a push priority adjustment sequence, the interactive learning load change set specifically refers to a response time abnormal task list, a correct rate decrease task list and a synchronous load offset identifier set, and the dynamically adjusted content difficulty configuration information includes a learning content difficulty grading index, a pushed content update mapping table and an adjusted task allocation list.

[0008] The improvements of the present invention are that the grammar error interaction recognition module includes: The response field extraction sub-module obtains the English learning interaction information of the user, extracts the error annotation field, the error type classification and the interaction timestamp in the response data according to the grammar task response data, arranges the data according to the interaction timestamp order, establishes the corresponding structure between the error type and the timestamp, and generates an error time annotation index set; The error density construction sub-module calls the error time annotation index set, divides all time periods into equal-length time windows according to the timestamp corresponding to the error type, counts the number of occurrences of the error type in each window, compares the difference in the cumulative number of errors in adjacent two time windows, calculates the cumulative change ratio of each error type, and obtains an error cumulative change ratio sequence; The error feature screening sub-module calls the error cumulative change ratio sequence, makes a difference judgment on the ratio value of each error type and the set error density change threshold, screens the error types whose change amplitude exceeds the error density change threshold, extracts the screened error types and the corresponding time series feature structures, and establishes a grammar interaction error feature list.

[0009] The improvement of the present invention is that the wrong item push frequency control module includes: The push content extraction sub-module obtains the grammar interaction error feature list, extracts the current push English learning content category corresponding to each error type, screens the push categories according to the error type field, establishes a push error category mapping relationship, and generates a push error category mapping table; The push frequency adjustment sub-module calls the push error category mapping table, and according to the error occurrence frequency, the error tolerance reference frequency, and the basic push frequency, and based on the numerical relationship between the parameters, screens the push content items whose frequencies need to be adjusted, and calculates the adjusted content push frequency; The content priority update sub-module calls the adjusted content push frequency, sorts the push content priorities according to the size of the push frequency, selects the content with an increased push frequency and adds it to the frequency increase list, selects the content with a decreased push frequency and adds it to the frequency decrease list, establishes a content increase and decrease frequency push list, and generates an interactive wrong item push frequency adjustment set.

[0010] The improvement of the present invention is that the learning load dynamic monitoring module includes: The interaction data extraction sub-module calls the interactive wrong item push frequency adjustment set, and according to the push task item, obtains the interactive response time and the answering correct rate identification data of the corresponding English learning content during the interaction process, establishes an interaction record basic table, and generates an interaction record basic data set; The synchronous change calculation sub-module calls the interaction record basic data set, extracts the response time and the answering correct rate of three consecutive tasks according to the task sequence, calculates the response time growth rate and the correct rate change rate of adjacent two tasks respectively, and calculates the task synchronous change load value; The learning load screening sub-module calls the task synchronous change load value, compares each task synchronous change load value with the set synchronous change threshold, screens the task set whose synchronous change load value exceeds the threshold, establishes a load abnormal task set, and generates an interactive learning load change set.

[0011] The improvement of the present invention is that the learning difficulty dynamic adjustment module includes: The task load ratio extraction sub-module calls the interactive learning load change set, extracts the current difficulty level information of the pushed English learning content, obtains the number of tasks with abnormal load and the total number of pushed tasks, calculates the proportion of abnormal tasks based on the number of tasks with abnormal load and the total number of pushed tasks, establishes the data of the proportion of abnormal tasks, and generates the data set of the proportion of abnormal tasks; The difficulty change judgment sub-module calls the data set of the proportion of abnormal tasks, compares the difference between the proportion value of abnormal tasks of the pushed tasks and the set load increase threshold and load decrease threshold, screens the task identification sets with the proportion of abnormal tasks higher than the load increase threshold and lower than the load decrease threshold, establishes the task difficulty change adjustment list, and generates the task difficulty change determination list; The difficulty configuration update sub-module calls the task difficulty change determination list, and according to the difficulty reduction identification set and the difficulty increase identification set, respectively lowers the current difficulty level of the content with the difficulty reduction identification by one level and raises the current difficulty level of the content with the difficulty increase identification by one level, adjusts the overall difficulty configuration of the pushed English learning content, and generates the dynamically adjusted content difficulty configuration information.

[0012] The improvement of the present invention is that the system further includes: The English skill interaction aggregation module calls the dynamically adjusted content difficulty configuration information, extracts the skill types corresponding to the categories of the pushed English learning content, including vocabulary, listening, grammar, and speaking, calculates the interaction contribution amount of each skill category according to the adjusted content push ratio of each skill type and combines the change value of the task interaction correct rate of each skill type of the user, summarizes the interaction contribution amount of each skill category, calculates the user's English skill score, and generates the user's English skill evaluation information; The user's English skill evaluation information is specifically the vocabulary skill interaction index, the listening skill interaction index, the grammar skill interaction index, and the speaking skill interaction index.

[0013] The improvement of the present invention is that the English skill interaction aggregation module includes: The skill category extraction sub-module calls the dynamically adjusted content difficulty configuration information, extracts the categories of the pushed English learning content, and classifies them into four skill types of vocabulary, listening, grammar, and speaking according to the category fields, forms the corresponding table of the pushed content and the skill types, and generates the skill category corresponding table; The skill contribution amount calculation sub-module calls the skill category corresponding table, extracts the content push ratio, the change value of the task interaction correct rate, the average difficulty coefficient of the content, and the response time volatility of each skill type, and calculates the interaction contribution amount of each skill category through operation, and generates the skill category contribution amount set; The English skill score generation sub-module calls the set of skill category contribution amounts, sums them up after weighting according to the skill category weight configuration standard based on the numerical value of each skill category contribution amount, generates the total English skill score of a single user, evaluates the online learning effect of the user's English skills, and generates the user's English skill evaluation information.

[0014] Compared with the prior art, the advantages and positive effects of the present invention are as follows: In the present invention, through the dynamic monitoring of the change in the error event density in the user's English learning interaction information, an error event time series is established by using the error annotation field, error type classification, and time stamp. The object of adjusting the push frequency of learning content is accurately determined according to the change range of the error density, so as to achieve efficient error item precise intensive training. At the same time, according to the synchronous change range of the task response time growth rate and the answer correct rate change rate, the change of the user's learning load is identified in real time, and the difficulty level of the learning content is automatically adapted dynamically, forming a benign dynamic regulation mechanism between the learning difficulty and the user's learning state, significantly improving the user's learning acceptance and effect. By classifying and calculating the interaction contribution amount and the change in interaction correct rate of each English skill category, the evaluation quantification of each English skill is accurately realized, effectively improving the objectivity and comprehensiveness of the learning effect evaluation. Brief Description of the Drawings

[0015] Figure 1 is the system flow chart of the present invention; Figure 2 is the flow chart of the grammar error interaction recognition module of the present invention; Figure 3 is the flow chart of the wrong item push frequency regulation module of the present invention; Figure 4 is the flow chart of the learning load dynamic monitoring module of the present invention; Figure 5 is the flow chart of the learning difficulty dynamic adjustment module of the present invention; Figure 6 is the flow chart of the English skill interaction aggregation module of the present invention. Detailed Embodiment

[0016] In order to make the purpose, technical solution and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the drawings and embodiments. It should be understood that the specific embodiments described here are only used to explain the present invention, and are not used to limit the present invention.

[0017] In the description of the present invention, it should be understood that the orientation or positional relationship indicated by the terms "length", "width", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc. is based on the orientation or positional relationship shown in the drawings. It is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be construed as a limitation to the present invention. In addition, in the description of the present invention, the meaning of "a plurality of" is two or more, unless otherwise specifically defined.

[0018] Please refer to Figure 1 , the present invention provides a technical solution: a mobile English learning system based on interaction, the system includes: The grammar error interaction recognition module obtains the English learning interaction information of the user. According to the grammar task response data, it extracts the error annotation field, error type classification, and interaction timestamp in the response data, constructs an error event time series using the error type field and the interaction timestamp, and based on the cumulative value of the error occurrence frequency within a preset time window, compares the cumulative value differences between two adjacent time windows, filters out the set of error types whose density change amplitude exceeds the set error density change threshold, and establishes a grammar interaction error feature list; The error event time series refers to an ordered data set commonly used in the fields of natural language processing and interactive learning to record the number of occurrences of specific types of errors at different time points, retrieved from the NLP field event time series modeling standard; the error density change threshold is a preset percentage change amount, which measures whether the error frequency change is abnormal within a certain time period; The wrong item push frequency regulation module calls the grammar interaction error feature list, extracts the current push English learning content category corresponding to each error type in the list, adjusts the push frequency of the English learning content corresponding to the error according to the error occurrence frequency and the original set push reference frequency, rearranges the push order and updates the content priority, filters out the sets of content that need to increase and decrease the frequency, and generates an interactive wrong item push frequency adjustment set; The learning load dynamic monitoring module calls the interactive wrong item push frequency adjustment set, obtains the interactive response time and the answer correct rate identifier recorded during the interaction of the English learning content according to the task items of the English learning content push, calculates the synchronous change amplitude based on the response time growth rate and the correct rate change rate of three consecutive tasks, filters out the set of tasks whose synchronous change amplitude exceeds the set synchronous change threshold, classifies them into the learning load abnormal task set, and generates an interactive learning load change set; The interactive response time refers to the time required for a user to complete an interactive task; the answer accuracy rate identifier refers to the annotation result of whether a single answer is correct (1 for correct, 0 for incorrect); the response time growth rate and the accuracy rate change rate refer to the growth rate per unit time and the proportion of the decrease in the accuracy rate in consecutive tasks, which are commonly found in the standards for dynamic analysis of user behavior; The learning difficulty dynamic adjustment module calls the set of changes in interactive learning load, extracts the current difficulty level of the English learning content to be pushed, judges the overall learning load status based on the proportion of the number of tasks with abnormal learning load, and determines the change range according to the load status and the set standard value for adjusting the learning difficulty. If the proportion of abnormal load exceeds the load increase threshold, the difficulty of the corresponding content is reduced by one level. If the proportion of abnormal load is lower than the load decrease threshold, the difficulty of the corresponding content is increased by one level, adjusts the difficulty configuration of the pushed English learning content, and generates the dynamically adjusted content difficulty configuration information; The standard value for adjusting the learning difficulty is the benchmark for setting the adjustment conditions. For example, when the proportion of abnormal tasks reaches 30%, the adjustment is triggered, and it is retrieved from the adjustment mechanism of the adaptive learning system; The English skill interactive aggregation module calls the dynamically adjusted content difficulty configuration information, extracts the skill types corresponding to the categories of the pushed English learning content, including vocabulary, listening, grammar, and speaking. According to the proportion of the pushed content of each skill type after adjustment, combined with the change value of the task interaction accuracy rate of each skill type of the user, calculates the interactive contribution amount of the skill category, sums up the interactive contribution amount of each skill category, calculates the user's English skill score, and generates the user's English skill evaluation information; The change value of the task interaction accuracy rate refers to the difference in the accuracy rate of the user's corresponding tasks before and after adjustment in each skill category; the weight allocation standard for skill categories is generally set according to different learning objectives. For example, listening and speaking each account for 30%, and vocabulary and grammar each account for 20%, which is retrieved from the comprehensive evaluation system for foreign language learning.

[0019] The list of grammar interaction error characteristics includes the density distribution of error types, the time sequence trajectory of error events, and the categories of error high-frequency words. The interactive wrong item push frequency adjustment set specifically includes the frequency supplement task set, the frequency reduction task set, and the push priority adjustment sequence. The set of changes in interactive learning load specifically refers to the list of tasks with abnormal response time, the list of tasks with decreased accuracy rate, and the set of synchronous load offset identifiers. The dynamically adjusted content difficulty configuration information includes the learning content difficulty classification index, the push content update mapping table, and the adjusted task allocation list. The user's English skill evaluation information specifically includes the vocabulary skill interaction index, the listening skill interaction index, the grammar skill interaction index, and the speaking skill interaction index.

[0020] Please refer to Figure 2 , the grammar error interaction recognition module includes a response field extraction sub-module, an error density construction sub-module, and an error feature screening sub-module; The response field extraction sub-module obtains the user's English learning interaction information. Based on the grammar task response data, it extracts the error annotation fields, error type classifications, and interaction timestamps in the response data, arranges the data in the order of interaction timestamps, establishes the corresponding structure between the error types and timestamps, and generates an error time annotation index set. Obtain the grammar task response data when the user performs grammar tasks in the mobile English learning system. First, collect the current grammar task interaction log data, and extract the error annotation fields, error type classification information, and interaction timestamp parameters in each interaction record. The error annotation field records whether the task is answered incorrectly. The error type classification classifies the error into specific grammar categories, such as incorrect use of articles, incorrect use of tenses, etc. The interaction timestamp records the system time at each task submission. For example, if it is set that the user completes an answer at 10:00 on May 1, 2025, and completes the next answer at 10:05 on May 1, 2025, the corresponding extracted timestamps are 1682906400 seconds and 1682906700 seconds respectively. After extraction, all the data is sorted in ascending order according to the interaction timestamps, so as to establish a correct time-ordered data chain. During the sorting process, the bubble sorting method is used to exchange adjacent data. Set the initial error type and timestamp data array as [(tense error, 1682906700), (article error, 1682906400)]. After one exchange, the order is adjusted to [(article error, 1682906400), (tense error, 1682906700)]. Subsequently, establish a one-to-one correspondence between the error types and timestamps, construct an error event record chain, for example, generate a mapping structure: article error -> 1682906400 seconds, tense error -> 1682906700 seconds, and then generate an error time annotation index set.

[0021] The error density construction sub-module calls the error time annotation index set. According to the timestamps corresponding to the error types, it divides all time periods into equal-length time windows, counts the number of occurrences of each error type within each window, compares the difference in the cumulative number of errors between adjacent two time windows, calculates the cumulative change ratio of each error type, and obtains a sequence of error cumulative change ratios. Call the error time annotation index set. First, extract a series of timestamp information corresponding to each error type, divide the time period into equal-length segments, set the time window width to 300 seconds, that is, 5 minutes. Divide the interval from 1682906400 to 1682906700 seconds into 1 window, and count the number of specific error types that appear in each time window. If there are 2 article errors and 0 tense errors in the first window 1682906400 - 1682906700, then the count of article errors is 2 and the count of tense errors is 0. Subsequently, perform a difference operation on the cumulative error counts in two adjacent time windows. For example, if there are 2 article errors in the first window and 3 article errors in the next window, the difference is 1. The cumulative change ratio is calculated according to the formula as the change amount divided by the error count in the previous window, that is, 1 / 2 = 0.5. To further unify the processing rules, for windows with a cumulative error count of 0, the difference is uniformly set to 0 to avoid a zero denominator situation. See Table 1 for details. Table 1 lists the statistical results of each error type and the cumulative change ratio in the example window. Finally, form an independent cumulative change ratio sequence for each error type to obtain the error cumulative change ratio sequence. Among them, the setting of the error density change threshold is based on the average error volatility statistically obtained from the user interaction dataset under the standard load of the system. Specifically, 100 users' daily interaction data for 7 days in the mobile English learning system are extracted, and the change ratio of the number of various grammar errors is statistically segmented every day. A total of 700 groups of data are obtained. The average value of the change ratios of each group of data is used as the benchmark. If the average change ratio is 0.28, the threshold is rounded up to 0.3 to ensure adaptation to the interaction habits of most users and avoid over-screening or missed selection. For example, in an actual test, the change ratios of article errors are 0.25, 0.32, 0.28, 0.29, 0.27 respectively, and the calculated average value is 0.282. The threshold is finally set to 0.3. The threshold fluctuates with the change trend of the average learning load of the user group. If the overall system load increases (such as a daily task volume increase of more than 20%), the threshold is appropriately increased to 0.35 to adapt to the systematic increase in the overall error rate.

[0022] Table 1 Statistical Table of Error Type Windows

[0023] Error type Window number Number of errors (times) Cumulative change ratio Article error 1 2 0.5 Tense error 1 0 0

[0024] Table 1 lists the occurrence times and cumulative change ratios of article errors and tense errors in two consecutive time windows, and the volatility of article errors between different windows can be observed.

[0025] The error feature screening sub-module calls the error cumulative change ratio sequence, makes a difference judgment on the ratio value of each error type and the set error density change threshold, screens out the error types whose change range exceeds the error density change threshold, extracts the screened error types and the corresponding time series feature structures, and establishes a grammar interaction error feature list; Call the error cumulative change ratio sequence, make a difference judgment on the ratio value of each error type and the set error density change threshold. Assume that the error density change threshold is set to 0.3. After analyzing the error change rate data in the interaction records of different users in the system within a day, the value range is concentrated between 0.2 and 0.4. Finally, the median value of 0.3 is selected as the standard. The specific setting process is as follows: collect 100 pieces of user interaction data, calculate the change rate array, sort it and take the median. If the change ratio of article errors from the first window to the second window is 0.5, then 0.5 minus 0.3 gets 0.2, which is greater than 0, indicating that the change range exceeds the set threshold. Therefore, article errors are screened out. On the contrary, if the change rate of a certain error type is 0.2, it is screened out. In the refined screening process, for each error type and its cumulative change ratio, perform an absolute value operation, compare its numerical size with 0.3 to ensure accurate screening. After screening, extract the names of the error types that pass the screening and their corresponding time series features, and finally establish a grammar interaction error feature list. Among them, the rationality of the error density change threshold comes from the statistical data of user interaction behavior fluctuations. In the test of expanding the sample size to 1000 task interactions, setting the threshold to 0.3 can screen out more than 85% of the error types with actual learning bottleneck characteristics, and at the same time avoid screening out the error types within the normal fluctuation range. If the subsequent system monitors that the overall error rate decreases by more than 10%, the threshold can be automatically adjusted down to 0.25 for dynamic adaptation.

[0026] Please refer to Figure 3 , the wrong item push frequency regulation module includes a push content extraction sub-module, a push frequency adjustment sub-module, and a content priority update sub-module; The push content extraction sub-module obtains the grammar interaction error feature list, extracts the current push English learning content category corresponding to each error type, screens the push category according to the error type field, establishes a push error category mapping relationship, and generates a push error category mapping table; Obtain a list of grammar interaction error features, extract the current pushed English learning content category corresponding to each error type. For each error type record, extract its error type field, such as extracting article errors, tense errors, subject-verb agreement errors, etc. During the data extraction process, adopt a traversal processing method to traverse all records in the grammar interaction error feature list and extract the corresponding pushed category field. The pushed category field is usually the identifier of a specific course module or learning unit, such as the basic grammar unit, the tense special training unit, etc. For example, if the error type is an article error, its corresponding pushed category is the "primary grammar module"; if the error type is a tense error, the corresponding pushed category is the "verb tense special training". Filter according to the extracted error type field, and eliminate error type records without a clear corresponding relationship with the pushed content, such as excluding unclassified errors, unknown errors, etc. records. During the filtering process, adopt a conditional judgment method. If the pushed category field is a null value, then eliminate the record; otherwise, keep it. After the filtering is completed, establish a mapping relationship between the pushed error categories to form a one-to-one mapping from the error type to the pushed content category. For example, the article error is mapped to the primary grammar module, and the tense error is mapped to the verb tense special training. Finally, form a pushed error category mapping table, as shown in Table 2. When filtering the pushed category, uniformly mark the error types that cannot be clearly attributed as "other" and do not participate in the push frequency adjustment operation in the subsequent processing, and generate a pushed error category mapping table.

[0027] Table 2 Pushed Error Category Mapping Table

[0028] Error type Push content category Article error Elementary grammar module Tense error Special training on verb tenses Subject-verb agreement error Special training on subject-verb agreement

[0029] Table 2 lists the mapping relationship between different grammar error types and their corresponding pushed English learning content categories, establishing a data foundation.

[0030] The push frequency adjustment sub-module calls the pushed error category mapping table. According to the error occurrence frequency, the error tolerance benchmark frequency, and the basic push frequency, and based on the numerical relationship between the parameters, filter the push content items whose frequencies need to be adjusted, using the formula: ; Calculate to obtain the adjusted content push frequency; Among them, represents the adjusted content push frequency, represents the error occurrence frequency, represents the error tolerance benchmark frequency, represents the basic push frequency, represents the average interval of error occurrences in consecutive interaction tasks, represents the benchmark error time interval; Call the push error category mapping table, and extract three frequency parameters according to the error occurrence frequency, error tolerance benchmark frequency, and basic push frequency corresponding to each push content. The error occurrence frequency represents the number of times this type of error is detected per unit time. For example, the error occurrence frequency of article errors in a certain push content is 0.15 times per minute. The error tolerance benchmark frequency is set as the maximum error frequency allowed by the system. The system is set to 0.1 times per minute based on past interaction samples. The specific setting basis is to extract the interaction data under normal learning load in the past 30 days, calculate the mean and standard deviation of the error frequency, and set the mean plus one standard deviation as the benchmark frequency. Therefore, if the mean is 0.07 times per minute and the standard deviation is 0.03 times per minute, the benchmark frequency is 0.1 times per minute. The basic push frequency is set as the standard push frequency. For example, if the default is to push 6 times per hour for each module, then is 0.1 times per minute. After extracting the three frequency parameters, use the formula: ; for calculation. Among them, is set as the standard error time interval. Based on a large number of interaction records in the system, the standard time interval is determined to be 600 seconds, that is, an error occurs on average every 10 minutes. The value is calculated from the average error submission interval time. Select the data of 300 users, record the daily interaction time, extract the time intervals of the first 50 error submissions, and take the average as the benchmark. If the average is statistically 598 seconds, round it up and set it to 600 seconds. is the average interval of error occurrences in continuous interaction tasks. For example, if the error submission times of a user in three consecutive tasks are 200 seconds, 500 seconds, and 800 seconds respectively, then is (500 - 200) + (800 - 500) = 600 seconds. Divide by 2 intervals, and the average value is 300 seconds. Substitute it into the formula and set , , , , , then calculate: ; ; ; Get the adjusted content push frequency of 0.7993, and then reclassify and sort the push content based on the size of the adjustment value of each content item, and finally form a set of push content frequency adjustment values.

[0031] The content priority update sub-module calls the adjusted content push frequency, sorts the push content priorities according to the size of the push frequency, selects the content with an increased push frequency and adds it to the frequency increase list, selects the content with a decreased push frequency and adds it to the frequency decrease list, establishes a content increase and decrease push list, and generates an interactive error item push frequency adjustment set; Call the adjusted content push frequency, sort according to the size of the adjustment value of each push content, extract the push content items with an increased adjustment value and add them to the frequency increase list, extract the push content items with a decreased adjustment value and add them to the frequency decrease list. During the sorting process, the adjustment values are arranged in descending order. The push content with a larger value has a higher priority, and the push content with a smaller value has a lower priority. For example, in the adjustment value set, article error corresponds to 0.7993, tense error corresponds to 0.6012, and subject-verb agreement error corresponds to 0.4125. After sorting from large to small by the adjustment value, they are article error, tense error, and subject-verb agreement error in turn. Set the threshold limit to 10% of the system push adjustment ratio, that is, add the content in the top 10% of the adjustment values to the frequency increase list, and add the content in the last 10% to the frequency decrease list. If the total number of push content items is 30, then take the first three and the last three for classification processing. The basis for setting this push adjustment ratio standard is the system push load test data. On the premise of limiting the daily push to no more than 60 pieces of content, controlling the adjustment ratio within 10% can maintain the balance between the stability of the push frequency and personalized adjustment. If the subsequent system monitors that the push response rate drops by more than 5%, then appropriately increase the push adjustment ratio to 15%. After finally updating the push list, establish a new push content priority configuration table and generate an interactive error item push frequency adjustment set.

[0032] Please refer to Figure 4 , the learning load dynamic monitoring module includes an interactive data extraction sub-module, a synchronous change calculation sub-module, and a learning load screening sub-module; The interactive data extraction sub-module calls the interactive error item push frequency adjustment set, and according to the push task items, obtains the interactive response time and the answering correct rate identification data of the corresponding English learning content during the interaction process, establishes an interactive record basic table, and generates an interactive record basic data set; Call the interaction error item push frequency adjustment set. According to the push task items, extract the interaction response time data and the answer correctness identification data of the corresponding English learning content during the interaction one by one. The response time data refers to the response time of each user after the push task is triggered, recorded in seconds. The answer correctness identification is of boolean type, where 1 represents a correct answer and 0 represents a wrong answer. During the data extraction process, use the task serial number as the index and record each interaction event in chronological order. Suppose the user executed task A, task B, and task C, and their interaction response times were 120 seconds, 150 seconds, and 180 seconds respectively, and the answer correct rates were 1, 0, and 1 respectively. The data is recorded as {task A: (120, 1), task B: (150, 0), task C: (180, 1)}. After extraction, establish a basic interaction record table, with task number, response time, and answer correctness as columns, and enter the records item by item. The sample of the basic table is shown in Table 3. When establishing the basic data table, if there is a task response timeout (such as the single task response time exceeds the system-set 1800 seconds), it is regarded as an invalid record and excluded to avoid abnormal data affecting subsequent calculations. The system default maximum response time is set based on the 95th percentile of the statistical results of the daily average interaction response duration of the actual user group, specifically extracted from the daily average interaction data. If the 95th percentile is 1750 seconds, the upper limit is set to 1800 seconds, and records exceeding the upper limit are excluded. Finally, organize and generate the basic interaction record data set.

[0033] Table 3 Basic Interaction Record Data Table

[0034] Task number Response time (seconds) Answer accuracy rate A 120 1 B 150 0 C 180 1

[0035] Table 3 lists the response times and answer correctness rates of three interaction tasks, providing basic data support for subsequent synchronous change calculations.

[0036] The synchronous change calculation sub-module calls the basic interaction record data set, extracts the response times and answer correctness rates of three consecutive tasks according to the task sequence, and calculates the response time growth rate and the correctness rate change rate of adjacent two tasks respectively. Use the formula: ; Calculate to obtain the task synchronous change load value; Among them, represents the task synchronous change load value, represents the continuous task response time growth rate, represents the continuous task answer correctness rate change rate; Call the basic interaction record data set, extract the response time and answer correctness rate data of three consecutive tasks according to the task sequence, and calculate the response time growth rate of adjacent two tasks and the correctness rate change rate , the response time growth rate is calculated by dividing the difference between the response time of the latter task and the response time of the former task by the response time of the former task. For example, the growth rate from task A to task B is (150 - 120) / 120 = 0.25, and from task B to task C is (180 - 150) / 150 = 0.2. The change rate of the answer correct rate is calculated by subtracting the correct rate label of the former task from the correct rate label of the latter task. For example, the change rate from task A to task B is (0 - 1) = -1, and from task B to task C is (1 - 0) = 1. Subsequently, the corrected formula: ; is used to calculate the synchronous change load value. Substitute the example data and set , , then the calculated absolute difference is , the absolute sum is , the ratio part is 1.25 / (1 + 0.75) = 0.7143, calculate the logarithmic part , that is, (0.25 / 2) + 1 = 1.125, , the final synchronous change load value is: ; After the calculation of the synchronous change load value is completed, the synchronous load values for each continuous three-task interaction are obtained, forming a set of task synchronous change load values. After adopting the new formula, the logarithmic part introduces the proportional change of relative to , enhancing the non-linear measurement of the impact of interaction delay changes, and thus more sensitively reflecting the abnormal characteristics of the interaction rhythm changes in synchronous changes.

[0037] The learning load screening sub-module calls the task synchronous change load value, compares each task synchronous change load value with the set synchronous change threshold, screens the set of tasks whose synchronous change load values exceed the threshold, establishes a set of tasks with abnormal load, and generates a set of interactive learning load changes; Call the task to synchronize the change load value, compare the synchronized change load value corresponding to each task with the set synchronization change threshold, filter out the tasks whose synchronized change load value exceeds the threshold, establish a set of tasks with abnormal load, and set the synchronization change threshold to 0.5 during the filtering process. The basis for setting this threshold is the standard of normal interactive load change of the system. Specifically, 1000 daily task interaction data are selected, the distribution of the synchronized change load value is calculated, and the load value corresponding to the 85th percentile is taken as the screening benchmark. If the value corresponding to the 85th percentile calculated is 0.47, it is rounded up to 0.5 to ensure that the screening results can capture most of the abnormally high-load tasks and will not introduce excessive noise tasks. For example, if the synchronized change load value of task A is 0.4, which is less than 0.5, it is not screened; if the synchronized change load value of task B is 0.7, which is greater than 0.5, it is screened into the abnormal set. All tasks are traversed in turn to complete the screening operation, and finally a set of tasks with abnormal load is established to generate an interactive learning load change set.

[0038] Please refer to Figure 5 , the learning difficulty dynamic adjustment module includes a task load ratio extraction sub-module, a difficulty change judgment sub-module, and a difficulty configuration update sub-module; The task load ratio extraction sub-module calls the interactive learning load change set, extracts the current difficulty level information of the pushed English learning content, obtains the number of tasks with abnormal load and the total number of pushed tasks, calculates the proportion of abnormal tasks based on the number of tasks with abnormal load and the total number of pushed tasks, establishes data on the proportion of abnormal tasks, and generates a data set on the proportion of abnormal tasks; Call the interactive learning load change set to extract the current difficulty level information of the pushed English learning content. The difficulty level information is usually marked as elementary, intermediate, or advanced for subsequent dynamic adjustment according to the learning load. During the data extraction process, the corresponding difficulty field is extracted by associating with the push task item number to establish a mapping relationship from the task number to the difficulty level. Subsequently, obtain the number of tasks with abnormal load and the total number of pushed tasks for each push task. The number of tasks with abnormal load is the number of tasks whose synchronized change load value exceeds the set threshold screened in the previous step, and the total number of pushed tasks is the total number of tasks actually pushed to the user and having interaction records under the same push content. For example, in the push content "Elementary Tense Special Practice", the total number of pushed tasks is 50, and the number of tasks with abnormal load is 12. Then calculate the abnormal task ratio as 12 / 50 = 0.24, that is, 24%. Extract the number of tasks with abnormal load and the corresponding total number of pushes respectively for all push contents, and calculate the abnormal task ratio item by item. The recording format of the abnormal task ratio is {task number: abnormal ratio}. When calculating the abnormal ratio, to prevent calculation anomalies caused by the denominator being 0, if the total number of pushed tasks is 0, the system defaults the abnormal ratio to be set to 0 to avoid affecting subsequent difficulty adjustment judgments. The basis for this processing standard is the analysis of the platform task push log records. The actual task push success rate is higher than 99.5%, and the probability of push failure is extremely low. Therefore, setting the 0 ratio processing rule is fully reasonable. After calculation, establish a data table of abnormal task ratios to form a data set of abnormal task ratios, and finally generate a data set of abnormal task ratios.

[0039] The difficulty change judgment sub-module calls the data set of abnormal task ratios, compares the difference between the abnormal task ratio value of the push task and the set load increase threshold and load decrease threshold, screens the task identification sets with abnormal task ratios higher than the load increase threshold and lower than the load decrease threshold, establishes a task difficulty change adjustment list, and generates a task difficulty change determination list; Call the dataset of the proportion of abnormal tasks. Compare the proportion of abnormal tasks of each push task with the preset load increase threshold and load decrease threshold of the system. The load increase threshold is set to 0.3, that is, when the proportion of abnormal tasks exceeds 30%, it is considered that the learning load is too heavy and the difficulty needs to be lowered. The load decrease threshold is set to 0.1, that is, when the proportion of abnormal tasks is less than 10%, it is considered that the learning load is too light and the difficulty needs to be increased. The basis for setting the load increase and decrease thresholds is based on the statistical analysis of user learning behavior data. By extracting the push and interaction data of 500 users in a continuous learning cycle, the average load abnormality rate distribution at each difficulty level is statistically analyzed. If it is found that the corresponding obvious task completion rate increases and the learning stagnation rate rises above 30% of the abnormality rate, while the learning completion rate increases and the error rate shows a downward trend below 10%, so it is reasonably set to 0.3 and 0.1, and it is verified in actual tests that the adjusted push strategy brings an overall course completion rate increase of more than 5%. In the actual screening process, compare the abnormal proportion of each push task item by item. If the abnormal proportion is greater than 0.3, classify the corresponding task number into the difficulty reduction identification set. If the abnormal proportion is less than 0.1, classify the task number into the difficulty increase identification set. For example, if the abnormal proportion of push task A is 0.35 and the abnormal proportion of push task B is 0.08, then task A enters the reduction list and task B enters the increase list, forming a task difficulty change adjustment list, and finally generating a task difficulty change determination list.

[0040] The difficulty configuration update sub-module calls the task difficulty change determination list. According to the difficulty reduction identification set and the difficulty increase identification set, lower the current difficulty level of the content marked for difficulty reduction by one level and raise the current difficulty level of the content marked for difficulty increase by one level, adjust the overall difficulty configuration of the pushed English learning content, and generate the dynamically adjusted content difficulty configuration information; Call the task difficulty change judgment list. According to the difficulty reduction flag set and the difficulty increase flag set, dynamically adjust the difficulty level of the pushed English learning content. For the pushed content extracted from the difficulty reduction flag set, lower the current difficulty level by one level. For example, if it was originally intermediate, lower it to elementary; if it was originally advanced, lower it to intermediate. For the pushed content extracted from the difficulty increase flag set during the upward adjustment, raise the current difficulty level by one level. For example, if it was originally elementary, raise it to intermediate; if it was originally intermediate, raise it to advanced. During the adjustment process, set a protection mechanism for situations beyond the upper and lower limits. For example, when it is currently elementary and still needs to be lowered, or when it is currently advanced and still needs to be raised, maintain the original difficulty level unchanged. The basis for setting this protection mechanism is to avoid the phenomenon of difficulty jump and out-of-bounds, which may lead to an unbalanced learning experience. This standard is obtained through actual push tests. In the test, if the protection mechanism is not set, the boundary overshoot rate is as high as 4%. After setting the protection mechanism, it drops to 0.2%, ensuring the rationality and continuity of the pushed difficulty adjustment. The adjustment operation is executed batchwise with the task identifier as the index, and a new pushed content difficulty level list is generated after the update to complete the update of the content difficulty configuration, and finally generate the dynamically adjusted content difficulty configuration information.

[0041] Please refer to Figure 6 , the English skill interaction aggregation module includes a skill category extraction sub-module, a skill contribution calculation sub-module, and an English skill score generation sub-module; The skill category extraction sub-module calls the dynamically adjusted content difficulty configuration information, extracts the categories of the pushed English learning content, and classifies them into four skill types: vocabulary, listening, grammar, and speaking according to the category fields, forms a correspondence table between the pushed content and the skill types, and generates a skill category correspondence table; Call the dynamically adjusted content difficulty configuration information, extract the category information of each piece of pushed English learning content. The category field records the detailed category labels of the current pushed content, such as vocabulary special training, listening comprehension practice, basic grammar explanation, oral situation dialogue practice, etc. According to the content of the category field, classify the pushed content into four skill types: vocabulary, listening, grammar, and speaking. The classification standard is to perform keyword matching based on the keywords in the field. For example, if the field contains keywords such as vocabulary and words, it is classified into the vocabulary skill type; if it contains listening and listening and speaking training, it is classified into the listening skill type; if it contains grammar and sentence structure, it is classified into the grammar skill type; if it contains speaking and dialogue, it is classified into the speaking skill type. When classifying, use string containment matching. If the pushed content is labeled as "daily oral practice", it is recognized and classified into the speaking skill type; if the pushed content is labeled as "basic vocabulary memory", it is recognized and classified into the vocabulary skill type. After classification, establish a one-to-one correspondence table between the pushed content number and the skill type. For example, push number T001 corresponds to the vocabulary skill, and T002 corresponds to the listening skill, forming a skill category correspondence table. During the classification process, for the pushed content that cannot be classified into the four major skill types, such as the comprehensive test module, it is uniformly marked as "unclassified" and excluded in subsequent statistical processing to avoid interfering with the calculation of skill scores. The proportion of unclassified pushed content is controlled within 5% of the total push volume. This standard is based on the analysis of the category distribution of platform pushed content. Among the full-volume pushed content, the proportion of clearly classified content reaches more than 95%. Therefore, the impact of excluding unclassified content on the overall data structure can be ignored, and finally a skill category correspondence table is generated.

[0042] The skill contribution calculation sub-module calls the skill category correspondence table, extracts the content push ratio, task interaction correct rate change value, content average difficulty coefficient, and response time volatility of each skill type, and uses the formula: ; Calculate the interaction contribution of each skill category through operation, and generate a set of skill category contributions; Among them, represents the interaction contribution of the th skill category, represents the push content ratio corresponding to the th skill category, represents the change value of the task interaction correct rate of the th skill category, represents the content average difficulty coefficient of the th skill category, represents the response time volatility of the th skill category task; Call the skill category correspondence table, and for each skill type, extract its content push ratio 、task interaction correct rate change value , Average content difficulty coefficient , Response time volatility , Content push ratio The calculation method is the number of pushed contents of this skill type divided by the total number of pushed contents. For example, if the number of pushed vocabulary contents is 20 and the total number of pushed contents is 100, then , Change value of task interaction accuracy is the change in the user's answer accuracy between two consecutive measurements. For example, if the first accuracy is 0.6 and the second is 0.7, then , Average content difficulty coefficient comes from the numerical processing of the difficulty level to which the pushed content belongs. For example, beginner = 1, intermediate = 2, advanced = 3. If 70% of the pushed contents of the vocabulary skill are beginner level and 30% are intermediate level, then , Response time volatility represents the ratio of the standard deviation to the mean of consecutive interaction response times. For example, if the response time samples are 90 seconds, 100 seconds, and 110 seconds, the standard deviation is 8.16 seconds, and the mean is 100 seconds, then , Subsequently, use the formula: ; Perform the operation of interaction contribution amount, substitute the example data, and set , , , , then calculate: ; ; ; Finally, the interaction contribution amount of this skill category is obtained as 0.0442. By introducing the dual correction mechanism of average content difficulty and response time volatility, it is ensured that the contribution amount not only reflects the change in the user's answer correctness, but also comprehensively considers the difficulty of learning content and the stability of interaction behavior, improving the accuracy and rationality of the evaluation of skill contribution amount, forming a skill dimension score vector, and finally generating a set of skill category contribution amounts.

[0043] The English skill scoring generation sub-module calls the set of skill category contribution amounts, sums up after weighting according to the skill category weight configuration standard based on the contribution amount value of each skill category, generates the total English skill score of a single user, evaluates the online learning effect of the user's English skills, and generates the user's English skill evaluation information; Call the set of contribution amounts for skill categories. According to the contribution amount values of each skill category, calculate the sum after weighting them respectively according to the skill category weight standard configured by the system. The skill category weight standard is set according to the learning application scenario. For example, in the standard English proficiency assessment system, the weights of the four skills of vocabulary, listening, grammar, and speaking are set at 25%, 25%, 25%, and 25% respectively. If it is applied to the workplace English training scenario, it can be adjusted to 30% for listening, 30% for speaking, 20% for vocabulary, and 20% for grammar. The system defaults to the standard equal distribution setting, which is consistent with the weight standard of the large-scale English proficiency certification exam according to the skill evaluation weights, ensuring the authority and wide adaptability of the evaluation system. In the actual weighting process, multiply the contribution amount by the corresponding weight coefficient. For example, if the contribution amount of the vocabulary skill is 0.0442 and the weight is 25%, the weighted score is 0.0442×0.25 = 0.01105. Calculate the weighted scores of the four skills in turn and sum them up to obtain the total English skill score of a single user. If the weighted scores of the four skills are 0.01105, 0.01500, 0.01050, and 0.01200 respectively, the total score is 0.01105 + 0.01500 + 0.01050 + 0.01200 = 0.04855. Finally, form the skill score index of the user and generate the user's English skill assessment information.

[0044] The above is only the preferred embodiment of the present invention, and it is not intended to limit the present invention in other forms. Any person skilled in the art may use the disclosed technical content to make changes or modifications into equivalent embodiments with equivalent changes and apply them to other fields. However, as long as it does not depart from the technical solution content of the present invention, any simple modification, equivalent change, and modification made to the above embodiments based on the technical essence of the present invention still fall within the protection scope of the technical solution of the present invention.

Claims

1. An interactive-based mobile English learning system, characterized in that The system includes: The grammar error interaction recognition module obtains the English learning interaction information of the user, based on the cumulative value of the error occurrence frequency within a preset time window, compares the cumulative value differences of two adjacent time windows, screens the set of error types whose density change amplitude exceeds the set error density change threshold, and establishes a grammar interaction error feature list; The wrong item push frequency regulation module calls the grammar interaction error feature list, and adjusts the push frequency of the English learning content corresponding to the error according to the error occurrence frequency and the originally set push reference frequency, and generates an interactive wrong item push frequency adjustment set; The learning load dynamic monitoring module calls the interactive wrong item push frequency adjustment set, obtains the interactive response time and the answer correctness identification recorded during the interaction of the English learning content, and screens the set of tasks whose synchronous change amplitude exceeds the set synchronous change threshold according to the response time growth rate and the correctness change rate of three consecutive tasks, and generates an interactive learning load change set; The learning difficulty dynamic adjustment module calls the interactive learning load change set, and adjusts the difficulty configuration of the pushed English learning content according to the proportion of the number of learning load abnormal tasks, and generates the dynamically adjusted content difficulty configuration information.

2. The interactive-based mobile English learning system according to claim 1, wherein The grammar interaction error feature list includes the error type density distribution, the error event time series trajectory, and the error high-frequency word category. The interactive wrong item push frequency adjustment set is specifically a frequency supplement task set, a frequency reduction task set, and a push priority adjustment sequence. The interactive learning load change set specifically refers to a response time abnormal task list, a correctness rate decrease task list, and a synchronous load offset identification set. The dynamically adjusted content difficulty configuration information includes a learning content difficulty classification index, a pushed content update mapping table, and an adjusted task allocation list.

3. The interactive-based mobile English learning system according to claim 2, wherein The grammar error interaction recognition module includes: The response field extraction sub-module obtains the English learning interaction information of the user, extracts the error annotation field, the error type classification, and the interaction timestamp in the response data according to the grammar task response data, arranges the data according to the interaction timestamp order, establishes the corresponding structure between the error type and the timestamp, and generates an error time annotation index set; The error density construction sub-module calls the error time annotation index set, divides all time periods into equal-length time windows according to the timestamps corresponding to the error types, counts the number of occurrences of each error type within each window, compares the difference in the cumulative number of errors between two adjacent time windows, calculates the cumulative change ratio of each error type, and obtains an error cumulative change ratio sequence; The error feature screening sub-module calls the error cumulative change ratio sequence, makes a difference judgment on the ratio value of each error type and the set error density change threshold, screens the error types whose change amplitude exceeds the error density change threshold, extracts the screened error types and the corresponding time series feature structures, and establishes a grammar interaction error feature list.

4. The interactive-based mobile English learning system according to claim 3, wherein The wrong item push frequency regulation module includes: The push content extraction sub-module obtains the list of grammar interaction error features, extracts the current push English learning content category corresponding to each error type, filters the push categories according to the error type field, establishes the mapping relationship of push error categories, and generates the push error category mapping table; The push frequency adjustment sub-module calls the push error category mapping table, and according to the error occurrence frequency, the error tolerance reference frequency, and the basic push frequency, filters the push content items whose frequencies need to be adjusted based on the numerical relationship between the parameters, and calculates the adjusted content push frequency; The content priority update sub-module calls the adjusted content push frequency, sorts the push content priorities according to the push frequency, selects the content with an increased push frequency and adds it to the frequency increase list, selects the content with a decreased push frequency and adds it to the frequency decrease list, establishes the content increase and decrease frequency push list, and generates the interaction error item push frequency adjustment set.

5. The interactive-based mobile English learning system according to claim 4, wherein The learning load dynamic monitoring module includes: The interaction data extraction sub-module calls the interaction error item push frequency adjustment set, obtains the interaction response time and the answer correct rate identification data of the corresponding English learning content during the interaction according to the push task item, establishes the basic interaction record table, and generates the basic interaction record data set; The synchronous change calculation sub-module calls the basic interaction record data set, extracts the response time and the answer correct rate of three consecutive tasks according to the task sequence, calculates the response time growth rate and the correct rate change rate of adjacent two tasks respectively, and calculates the task synchronous change load value; The learning load screening sub-module calls the task synchronous change load value, compares each task synchronous change load value with the set synchronous change threshold, screens the task set with the synchronous change load value exceeding the threshold, establishes the load abnormal task set, and generates the interactive learning load change set.

6. The interactive-based mobile English learning system according to claim 5, wherein, The learning difficulty dynamic adjustment module includes: The task load ratio extraction sub-module calls the interactive learning load change set, extracts the current difficulty level information of the push English learning content, obtains the number of load abnormal tasks and the total number of push tasks, calculates the abnormal task ratio according to the number of load abnormal tasks and the total number of push tasks, establishes the abnormal task ratio data, and generates the abnormal task ratio data set; The difficulty change judgment sub-module calls the abnormal task ratio data set, makes a difference comparison according to the abnormal task ratio value of the push task and the set load increase threshold and load decrease threshold, screens the task identification set with the abnormal task ratio higher than the load increase threshold and the task identification set lower than the load decrease threshold, establishes the task difficulty change adjustment list, and generates the task difficulty change determination list; The difficulty configuration update sub-module calls the task difficulty change determination list, according to the difficulty reduction identification set and the difficulty increase identification set, lowers the current difficulty level of the difficulty reduction identification content by one level and raises the current difficulty level of the difficulty increase identification content by one level respectively, adjusts the overall difficulty configuration of the push English learning content, and generates the dynamically adjusted content difficulty configuration information.

7. The interactive-based mobile English learning system according to claim 6, wherein, The system further includes: The English skill interaction aggregation module calls the dynamically adjusted content difficulty configuration information, extracts and pushes the skill types corresponding to the English learning content categories, including vocabulary, listening, grammar, and speaking. According to the adjusted content push ratio of each skill type and combined with the change value of the task interaction correct rate of each skill type of the user, it calculates the interaction contribution amount of each skill category, aggregates the interaction contribution amount of each skill category, calculates the user's English skill score, and generates the user's English skill evaluation information; The user's English skill evaluation information is specifically the vocabulary skill interaction index, the listening skill interaction index, the grammar skill interaction index, and the speaking skill interaction index.

8. The interactive-based mobile English learning system according to claim 7, wherein The English skill interaction aggregation module includes: The skill category extraction sub-module calls the dynamically adjusted content difficulty configuration information, extracts and pushes the English learning content categories, classifies them into four skill types of vocabulary, listening, grammar, and speaking according to the category fields, forms a corresponding table of the pushed content and the skill types, and generates a skill category corresponding table; The skill contribution amount calculation sub-module calls the skill category corresponding table, extracts the content push ratio, the change value of the task interaction correct rate, the average content difficulty coefficient, and the response time volatility of each skill type, and calculates the interaction contribution amount of each skill category through operations to generate a set of skill category contribution amounts; The English skill score generation sub-module calls the set of skill category contribution amounts, sums them up after weighting according to the skill category weight configuration standard based on the contribution amount value of each skill category, generates the total English skill score of a single user, evaluates the online learning effect of the user's English skills, and generates the user's English skill evaluation information.

Images