Personalized english education system and method based on ai multi-modal and big data

Abstract

The application relates to the technical field of intelligent education, and discloses an individualized English education system and method based on AI multi-modal and big data, which comprises the following steps: collecting learning data of learners in reading comprehension, listening recognition, vocabulary selection, follow-up reading, oral expression and writing application tasks; establishing a knowledge task mapping relationship between the same knowledge point and multiple task types; and generating a task performance vector; constructing a cross-task performance distribution based on the task performance vector; calculating a cross-task consistency index; generating an ability fault value according to the performance difference between input and output tasks, and determining the ability fault type; further generating a compensation training strategy, a learning priority and an individualized learning path, and outputting corresponding compensation training content. The application can correlate and analyze the learning performance of the same knowledge point in different tasks by establishing the knowledge task mapping relationship and generating the task performance vector, so that the comprehensiveness of the ability evaluation is improved.

Description

Technical Field

[0001] This invention relates to the field of intelligent education technology, and in particular to a personalized English education system and method based on AI multimodal and big data. Background Technology

[0002] With the development of artificial intelligence, big data, and online education technologies, English education is gradually shifting from a traditional, uniform teaching model to a data-driven, personalized teaching model.

[0003] Currently, some English education systems record learning data separately for reading comprehension, listening comprehension, vocabulary selection, oral expression, and writing tasks. They then generate ability evaluation results based on the accuracy, completion time, scoring, or error patterns of each task. For example, reading comprehension results evaluate reading ability, listening comprehension results evaluate listening ability, oral expression results evaluate oral ability, and writing results evaluate writing ability. The system then generates a learning report or recommends corresponding training content based on the evaluation results for each task.

[0004] However, in the process of realizing the technical solution of this application, the inventors of this application discovered that the prior art usually evaluates learners’ abilities based on the results of a single task, lacking a correlation analysis process between the same knowledge point and multiple task types. This makes it difficult to uniformly represent the learning performance of the same knowledge point in tasks such as reading, listening, speaking and writing, resulting in a one-sided problem in ability assessment results. Summary of the Invention

[0005] To address the above shortcomings, this invention provides a personalized English education system and method based on AI multimodal and big data, aiming to improve the existing technology where the learning performance of the same knowledge point in tasks such as reading, listening, speaking and writing is difficult to represent uniformly, resulting in one-sided ability assessment results.

[0006] In a first aspect, the present invention provides the following technical solution: a personalized English education system based on AI multimodal and big data, comprising:

[0007] The multimodal learning data acquisition module collects learners' learning data in reading comprehension tasks, listening recognition tasks, vocabulary selection tasks, shadowing tasks, oral expression tasks, and writing application tasks, and generates corresponding multimodal learning data.

[0008] The knowledge analysis module is connected to the multimodal learning data acquisition module. Based on the multimodal learning data, it identifies the corresponding knowledge points, establishes a knowledge-task mapping relationship between the same knowledge point and multiple task types, and generates task performance data and task performance vectors.

[0009] The cross-task consistency analysis module is connected to the knowledge analysis module. It constructs a cross-task performance distribution based on the task performance vector corresponding to the same knowledge point, calculates the corresponding cross-task consistency index, and generates knowledge point consistency analysis results.

[0010] The capability gap identification module is connected to the cross-task consistency analysis module. It generates capability gap values ​​based on the knowledge point consistency analysis results and determines the capability gap type based on the capability gap values ​​and differences in task performance.

[0011] The learning strategy generation module is connected to the capability gap identification module. It determines the training task category based on the capability gap type, determines the training task weight based on the capability gap value, constructs a set of compensation training requirements, generates a compensation training strategy based on the set of compensation training requirements, and generates a learning priority based on the compensation training strategy, knowledge point importance information, and knowledge point learning status information.

[0012] The education service module is connected to the learning strategy generation module, updates the personalized learning path according to the learning priority, and outputs the corresponding compensation training content.

[0013] Preferably, in the knowledge analysis module, the step of establishing the knowledge-task mapping relationship between the same knowledge point and multiple task types includes:

[0014] Extract knowledge point information from multimodal learning data;

[0015] Identify the task type corresponding to each knowledge point;

[0016] Establish a mapping relationship between the same knowledge point and multiple task types;

[0017] A knowledge task mapping relationship is generated based on the aforementioned association mapping relationship.

[0018] Preferably, in the knowledge analysis module, the steps of generating task performance data and task performance vectors include:

[0019] Obtain task completion result data, task completion time data, task error count data, and task quality score data for each knowledge point;

[0020] Standardize the task completion result data, task completion time data, task error count data, and task quality score data.

[0021] Generate task performance data based on the standardized processing results;

[0022] Construct a task performance vector based on the performance data of multiple tasks corresponding to the same knowledge point.

[0023] Preferably, in the cross-task consistency analysis module, the step of calculating the corresponding cross-task consistency index includes:

[0024] Obtain the task performance vector corresponding to the same knowledge point;

[0025] Construct a cross-task performance distribution based on the task performance vector;

[0026] Calculate the average performance value corresponding to the cross-task performance distribution;

[0027] Calculate the degree of dispersion between the performance data of each task and the average performance value;

[0028] A cross-task consistency index is generated based on the degree of dispersion.

[0029] Knowledge point consistency analysis results are generated based on the cross-task consistency index.

[0030] Preferably, in the capability fault identification module, the step of generating capability fault values ​​based on the knowledge point consistency analysis results includes:

[0031] Reading comprehension, listening recognition, and vocabulary selection tasks are categorized as input-type tasks.

[0032] The tasks of reading aloud, oral expression, and writing application are classified as output tasks.

[0033] Calculate the performance value of the input class task corresponding to the input class task;

[0034] Calculate the performance value of the output class task corresponding to the output class task;

[0035] Ability fault values ​​are generated based on the differences between the performance values ​​of input tasks and the performance values ​​of output tasks.

[0036] Preferably, in the capability fault identification module, the step of determining the type of capability fault includes:

[0037] Calculate the performance differences between reading comprehension and oral expression tasks;

[0038] Compute the performance differences between vocabulary selection tasks and writing application tasks;

[0039] Calculate the performance differences between the listening recognition task and the reading aloud task;

[0040] Identify the combination of tasks with the greatest performance differences;

[0041] The capability fault type is determined based on the maximum difference task combination and the capability fault value;

[0042] The types of ability gaps include oral output gaps, writing application gaps, listening input gaps, and comprehensive application gaps.

[0043] Preferably, in the learning strategy generation module, the step of generating a compensation training strategy based on the compensation training requirement set includes:

[0044] Identify the capability gap type; determine the training task category based on the capability gap type;

[0045] The weights of the training tasks are determined based on the capability gap values;

[0046] Construct a set of compensation training requirements based on training task categories and training task weights;

[0047] A compensation training strategy is generated based on the set of compensation training requirements.

[0048] Preferably, in the learning strategy generation module, the step of generating learning priorities includes:

[0049] Acquire compensatory training strategies; acquire information on the importance of knowledge points;

[0050] Obtain information on the learning status of knowledge points;

[0051] Learning priorities are generated based on the compensation training strategy, the importance of knowledge points, and the learning status of knowledge points; multiple knowledge points are then sorted according to the learning priorities.

[0052] Preferably, the step of determining the training task category based on the capability gap type includes:

[0053] When the type of ability gap is oral output gap, determine the situational question and answer training, oral retelling training and impromptu expression training;

[0054] When the type of competency gap is writing application gap, the training should include sentence construction training, sentence transformation training, and writing application training.

[0055] When the ability gap type is auditory input gap, determine auditory discrimination training, keyword localization training, and sound-meaning matching training;

[0056] When the competency gap type is a comprehensive application gap, determine the combined training of listening, speaking, reading and writing.

[0057] Secondly, this invention provides the following technical solution: a personalized English education method based on AI multimodal and big data, comprising the following steps:

[0058] S1. Collect learners' learning data in reading comprehension tasks, listening recognition tasks, vocabulary selection tasks, shadowing tasks, oral expression tasks, and writing application tasks, and generate multimodal learning data.

[0059] S2. Identify the corresponding knowledge points based on the multimodal learning data, establish a knowledge-task mapping relationship between the same knowledge point and multiple task types, and generate task performance data and task performance vectors.

[0060] S3. Construct cross-task performance distribution based on the task performance vector corresponding to the same knowledge point, calculate the cross-task consistency index, and generate knowledge point consistency analysis results.

[0061] S4. Calculate the performance values ​​of input tasks and output tasks based on the consistency analysis results of the knowledge points, and generate capability stratification values ​​based on the differences between the performance values ​​of input tasks and output tasks.

[0062] S5. Determine the type of capability gap based on the differences in task performance, the combination of tasks with the greatest differences, and the capability gap value.

[0063] S6. Determine the training task category based on the capability fault type, determine the training task weight based on the capability fault value, construct a set of compensation training requirements, and generate a compensation training strategy.

[0064] S7. Generate learning priorities based on the compensation training strategy, knowledge point importance information, and knowledge point learning status information, update personalized learning paths, and output corresponding compensation training content.

[0065] The present invention has the following beneficial effects:

[0066] 1. This invention establishes a knowledge-task mapping relationship between the same knowledge point and multiple task types, and generates a task performance vector, enabling the learning performance of the same knowledge point in tasks such as reading, listening, speaking, and writing to be analyzed in relation to each other, thereby reducing the bias in ability assessment that exists when relying solely on the results of a single task.

[0067] 2. This invention calculates a cross-task consistency index and generates capability gap values ​​based on the differences between the performance values ​​of input tasks and output tasks, enabling the identification of capability differences between the understanding and application stages of knowledge points, thereby improving the targeting of learning ability analysis.

[0068] 3. This invention generates a compensation training strategy based on the type and value of the ability gap and updates the personalized learning path, so that the training content can be matched with the learner's current ability status, thereby reducing repetitive training and improving the utilization efficiency of learning resources. Attached Figure Description

[0069] Figure 1 This is a diagram illustrating the architecture of the personalized English education system based on AI multimodal and big data proposed in this invention.

[0070] Figure 2 This is a schematic diagram illustrating the construction of the knowledge task mapping relationship proposed in this invention;

[0071] Figure 3 This is a flowchart of the cross-task consistency analysis proposed in this invention;

[0072] Figure 4 This is a flowchart of the capability tomography identification process proposed in this invention;

[0073] Figure 5 A flowchart for generating the learning strategy proposed in this invention;

[0074] Figure 6 This is a flowchart of the personalized English education method based on AI multimodal and big data proposed in this invention. Detailed Implementation

[0075] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0076] Example 1: Refer to Figure 1 In the first embodiment of the present invention, the present invention provides a personalized English education system based on AI multimodal and big data, comprising:

[0077] The multimodal learning data acquisition module collects learners' learning data in reading comprehension tasks, listening recognition tasks, vocabulary selection tasks, shadowing tasks, oral expression tasks, and writing application tasks, and generates corresponding multimodal learning data.

[0078] Specifically, the multimodal learning data acquisition module is deployed on the learning terminal or server side, including mobile terminals, tablet terminals, or computer terminals. This module receives learners' interaction data from various English learning tasks and stores it according to task type. In reading comprehension tasks, it collects the answers, answer times, and error records for each reading question; in listening comprehension tasks, it collects the listening comprehension answers, number of replays, and response time; in vocabulary selection tasks, it collects the vocabulary selection results and error records; in shadowing tasks, it collects audio data and shadowing score data; in oral expression tasks, it collects audio data, semantic matching score data, and expression completeness score data; and in writing application tasks, it collects text content, grammar score data, and vocabulary usage score data.

[0079] To achieve unified processing of data from different tasks, the multimodal learning data acquisition module performs task-identification association processing on the acquired raw data. Let the learner identifier be... The knowledge point is marked as The task type is identified as The corresponding learning record is represented as:

[0080] ;

[0081] in, This represents the data indicating the task completion result; This indicates the task completion time data; This data represents the number of task errors. This represents the task quality score data. For reading comprehension, listening recognition, and vocabulary selection tasks, the task quality score data is generated from the accuracy data of the corresponding tasks; for shadowing, oral expression, and writing application tasks, the task quality score data is generated from the scoring results output by the system.

[0082] Subsequently, the multimodal learning data acquisition module establishes a data index relationship based on learner identifiers, knowledge point identifiers, and task type identifiers, and aggregates learning records belonging to the same learner and corresponding to the same knowledge point to form a data set based on the knowledge point dimension:

[0083] ;

[0084] in, Indicate learner Corresponding knowledge points A collection of learning records; Representing knowledge points The number of associated task types.

[0085] After data collection is completed, the multimodal learning data acquisition module outputs the corresponding multimodal learning data to the knowledge analysis module for use in subsequent knowledge task mapping relationship construction, task performance data generation, and cross-task consistency analysis.

[0086] By using the above methods, learning data from reading comprehension tasks, listening recognition tasks, vocabulary selection tasks, shadowing tasks, oral expression tasks, and writing application tasks are uniformly collected into a dataset of the same knowledge point dimension. This enables subsequent processing to conduct analysis based on multi-task learning records corresponding to the same knowledge point, providing a data foundation for constructing knowledge task mapping relationships and generating task performance vectors.

[0087] Reference Figure 2Furthermore, the knowledge analysis module is connected to the multimodal learning data acquisition module. Based on the multimodal learning data, it identifies the corresponding knowledge points, establishes a knowledge-task mapping relationship between the same knowledge point and multiple task types, and generates task performance data and task performance vectors.

[0088] Preferably, in the knowledge analysis module, the steps for establishing a knowledge-task mapping relationship between the same knowledge point and multiple task types include:

[0089] Extract knowledge point information from multimodal learning data;

[0090] Identify the task type corresponding to each knowledge point;

[0091] Establish a mapping relationship between the same knowledge point and multiple task types;

[0092] Generate knowledge task mapping relationships based on the association mapping relationships.

[0093] Preferably, in the knowledge analysis module, the steps for generating task performance data and task performance vectors include:

[0094] Obtain task completion result data, task completion time data, task error count data, and task quality score data for each knowledge point;

[0095] Standardize the task completion result data, task completion time data, task error count data, and task quality score data.

[0096] Generate task performance data based on the standardized processing results;

[0097] Construct a task performance vector based on the performance data of multiple tasks corresponding to the same knowledge point.

[0098] Specifically, the knowledge analysis module receives multimodal learning data output from the multimodal learning data acquisition module and parses and processes the text content, speech content, and task identification information within the multimodal learning data. For words, phrases, sentence patterns, and semantic units involved in reading comprehension tasks, listening recognition tasks, vocabulary selection tasks, shadowing tasks, oral expression tasks, and writing application tasks, the knowledge analysis module matches and identifies them based on a pre-set knowledge base, assigning a corresponding knowledge point identifier to each knowledge point, thereby establishing a correspondence between knowledge points and learning records. Subsequently, a knowledge-task mapping relationship is established based on the knowledge point identifier and the task type identifier, forming an association structure where the same knowledge point corresponds to multiple task types.

[0099] Let the set of knowledge points be represented as: ;

[0100] The task type set is represented as: ;

[0101] in, Indicates the first One knowledge point; Indicates the first There are several task types. The knowledge analysis module establishes a knowledge-task mapping matrix based on the correspondence between knowledge points and task types: ;in, Representing knowledge points With task type The mapping state between them; when knowledge points Appears in task type China Times, ;otherwise, .

[0102] After constructing the knowledge-task mapping relationship, the knowledge analysis module further acquires the task completion result data, task completion time data, task error count data, and task quality score data corresponding to each knowledge point, and performs normalization processing on each to eliminate the impact of differences in the units of different task indicators on subsequent analysis. After normalization, the corresponding task performance data is generated. Let the knowledge point be... In task type The corresponding task performance data is represented as follows: Then, the task performance vector corresponding to the same knowledge point is represented as:

[0103] ;

[0104] in, Representing knowledge points The corresponding task performance vector; Representing knowledge points In task type The corresponding task performance data; Representing knowledge points The number of corresponding task types.

[0105] The knowledge analysis module sends the generated task performance vector to the cross-task consistency analysis module as the data basis for subsequent construction of cross-task performance distribution and calculation of cross-task consistency index.

[0106] The above method enables the unified association of learning records for the same knowledge point across reading comprehension, listening recognition, vocabulary selection, shadowing, oral expression, and writing application tasks, forming corresponding task performance vectors. This provides a unified data input for subsequent cross-task consistency analysis and ability gap identification.

[0107] Reference Figure 3Furthermore, the cross-task consistency analysis module is connected to the knowledge analysis module. It constructs a cross-task performance distribution based on the task performance vector corresponding to the same knowledge point, calculates the corresponding cross-task consistency index, and generates knowledge point consistency analysis results.

[0108] Preferably, in the cross-task consistency analysis module, the steps for calculating the corresponding cross-task consistency index include:

[0109] Obtain the task performance vector corresponding to the same knowledge point;

[0110] Construct a cross-task performance distribution based on the task performance vector;

[0111] Calculate the average performance value corresponding to the cross-task performance distribution;

[0112] Calculate the degree of dispersion between the performance data of each task and the average performance value;

[0113] Generate a cross-task consistency index based on the degree of dispersion;

[0114] Knowledge point consistency analysis results are generated based on the cross-task consistency index.

[0115] Specifically, the cross-task consistency analysis module receives the task performance vector output by the knowledge analysis module and performs statistical analysis on the task performance data of the same knowledge point across multiple task types, using knowledge points as the analysis object. For knowledge points... The corresponding task performance vector is represented as follows:

[0116] ;

[0117] in, Representing knowledge points The corresponding task performance vector; Representing knowledge points In task type The corresponding task performance data; Representing knowledge points The corresponding number of task types. The cross-task consistency analysis module first constructs a cross-task performance distribution based on the performance data of each task in the task performance vector, so as to represent the performance of the same knowledge point in reading comprehension tasks, listening recognition tasks, vocabulary selection tasks, shadowing tasks, oral expression tasks, and writing application tasks.

[0118] After constructing the cross-task performance distribution, the average performance value corresponding to each knowledge point is calculated:

[0119] ;

[0120] in, Representing knowledge points The corresponding average performance value; Representing knowledge points In task type The corresponding task performance data; This indicates the number of task types. The average performance value is used to characterize the overall performance level of the same knowledge point across multiple task types. Subsequently, the cross-task consistency analysis module calculates the degree of dispersion between the performance data of each task and the average performance value:

[0121] ;

[0122] in, Representing knowledge points The corresponding degree of dispersion; Representing knowledge points In the mission The corresponding task performance data; Representing knowledge points The corresponding average performance value. Dispersion is used to characterize the performance differences of the same knowledge point across different task types.

[0123] After obtaining the degree of dispersion, a cross-task consistency index is further generated:

[0124] ;

[0125] in, Representing knowledge points The corresponding cross-task consistency index; Representing knowledge points The corresponding degree of dispersion. The cross-task consistency index is used to characterize the consistency of the performance of the same knowledge point across multiple task types. When the task performance data corresponding to different task types are relatively similar, the dispersion decreases, and the corresponding cross-task consistency index increases; when the differences in task performance data corresponding to different task types increase, the dispersion increases, and the corresponding cross-task consistency index decreases. Subsequently, the cross-task consistency analysis module associates and stores the knowledge point identifier, the cross-task consistency index, and the corresponding cross-task performance distribution to form the knowledge point consistency analysis result, which is then sent to the capability fault identification module. The knowledge point consistency analysis result includes at least the knowledge point identifier, task performance vector, average performance value, dispersion, and cross-task consistency index.

[0126] The above methods enable unified analysis of the performance data of the same knowledge point across multiple task types, and the cross-task consistency index is used to characterize the consistency of the knowledge point's performance across different task types, thus providing a data foundation for the subsequent generation of capability gap values ​​and the determination of capability gap types.

[0127] Reference Figure 4 Furthermore, the capability gap identification module connects with the cross-task consistency analysis module to generate capability gap values ​​based on the knowledge point consistency analysis results, and determines the capability gap type based on the capability gap values ​​and differences in task performance.

[0128] Preferably, in the capability fault identification module, the steps for generating capability fault values ​​based on the knowledge point consistency analysis results include:

[0129] Reading comprehension, listening recognition, and vocabulary selection tasks are categorized as input-type tasks.

[0130] The tasks of reading aloud, oral expression, and writing application are classified as output tasks.

[0131] Calculate the performance value of the input class task corresponding to the input class task;

[0132] Calculate the performance value of the output class task corresponding to the output class task;

[0133] Ability fault values ​​are generated based on the differences between the performance values ​​of input tasks and the performance values ​​of output tasks.

[0134] Preferably, in the capability fault identification module, the steps for determining the type of capability fault include:

[0135] Calculate the performance differences between reading comprehension and oral expression tasks;

[0136] Compute the performance differences between vocabulary selection tasks and writing application tasks;

[0137] Calculate the performance differences between the listening recognition task and the reading aloud task;

[0138] Identify the combination of tasks with the greatest performance differences;

[0139] The capability gap type is determined based on the combination of the most different tasks and the capability gap value.

[0140] Among them, the types of ability gaps include oral output gaps, writing application gaps, listening input gaps, and integrated application gaps.

[0141] Specifically, the competency gap identification module receives the knowledge point consistency analysis results output by the cross-task consistency analysis module, and based on the task performance data of the same knowledge point in different task types, performs a difference analysis on the comprehension and application abilities corresponding to the knowledge point. First, task types are classified according to task attributes, with reading comprehension, listening recognition, and vocabulary selection tasks classified as input tasks, and shadowing, oral expression, and writing application tasks classified as output tasks. For knowledge points... The task performance data corresponding to the input task and the task performance data corresponding to the output task are obtained respectively, and the corresponding input task performance value and output task performance value are calculated.

[0142] The performance values ​​of input-type tasks are represented as follows:

[0143] ;

[0144] in, Representing knowledge points The corresponding input task performance value; Representing knowledge points Task performance data in the kth input class task; This indicates the number of input tasks.

[0145] The performance values ​​of output tasks are represented as follows:

[0146] ;

[0147] in, Representing knowledge points The corresponding output task performance value; Representing knowledge points Task performance data in the kth output class task; Indicates the number of output tasks.

[0148] Subsequently, capability gradation values ​​are generated based on the differences between input task performance values ​​and output task performance values: ;in, Representing knowledge points The corresponding capability fault value; Indicates the performance value of the input task; This represents the performance value for the output type of task. The ability stratification value is used to characterize the performance difference of the same knowledge point between comprehension-type tasks and application-type tasks.

[0149] After obtaining the competency gap values, the competency gap identification module further calculates the performance differences between reading comprehension and oral expression tasks, vocabulary selection and writing application tasks, and listening recognition and shadowing tasks. Let the task performance differences be represented as:

[0150] ;

[0151] in, Indicates differences in task performance; Representing knowledge points In task type The corresponding task performance data; Representing knowledge points In task type The corresponding task performance data is then analyzed. The capability gap identification module compares the differences in task performance for each task combination and identifies the task combination with the largest difference.

[0152] When the task combination with the greatest difference corresponds to the reading comprehension task and the oral expression task, it is determined as the oral output gap by combining the ability gap value; when the task combination with the greatest difference corresponds to the vocabulary selection task and the writing application task, it is determined as the writing application gap by combining the ability gap value.

[0153] When the task combination with the greatest difference corresponds to both the listening recognition task and the reading aloud task, it is identified as the listening input segment based on the ability segment value; when there are significant differences in multiple task combinations, it is identified as the comprehensive application segment. Subsequently, the ability segment identification module associates and stores the ability segment value, the task combination with the greatest difference, and the ability segment type, and sends them to the learning strategy generation module.

[0154] The above method can generate capability fault values ​​by utilizing the difference between the performance values ​​of input tasks and output tasks, and determine the capability fault type by combining the task combination with the largest difference. This enables the identification of the application difference state of the same knowledge point in different task types, providing a data foundation for the construction of subsequent compensation training requirements.

[0155] Reference Figure 5 Furthermore, the learning strategy generation module connects with the capability gap identification module to determine the training task category based on the capability gap type, determine the training task weight based on the capability gap value, construct a set of compensation training requirements, generate a compensation training strategy based on the set of compensation training requirements, and generate a learning priority based on the compensation training strategy, the importance information of knowledge points, and the learning status information of knowledge points.

[0156] Preferably, in the learning strategy generation module, the steps of generating a compensation training strategy based on the compensation training requirement set include:

[0157] Identify the capability gap type; determine the training task category based on the capability gap type;

[0158] The weights of the training tasks are determined based on the capability gap values;

[0159] Construct a set of compensation training requirements based on training task categories and training task weights;

[0160] A compensation training strategy is generated based on the set of compensation training requirements.

[0161] Preferably, the steps for determining the training task category based on the type of capability gap include:

[0162] When the type of ability gap is oral output gap, determine the situational question and answer training, oral retelling training and impromptu expression training;

[0163] When the type of competency gap is writing application gap, the training should include sentence construction training, sentence transformation training, and writing application training.

[0164] When the ability gap type is auditory input gap, determine auditory discrimination training, keyword localization training, and sound-meaning matching training;

[0165] When the competency gap type is a comprehensive application gap, determine the combined training of listening, speaking, reading and writing.

[0166] Preferably, in the learning strategy generation module, the steps for generating learning priorities include:

[0167] Acquire compensatory training strategies; acquire information on the importance of knowledge points;

[0168] Obtain information on the learning status of knowledge points;

[0169] Learning priorities are generated based on the compensation training strategy, the importance of knowledge points, and the learning status of knowledge points; multiple knowledge points are then sorted according to the learning priorities.

[0170] Specifically, the learning strategy generation module receives the ability gap identification module's output ability gap values ​​and types, and generates training task categories based on the ability differences corresponding to the ability gap types. For the oral output gap, training task categories include situational question-and-answer training, oral retelling training, and impromptu expression training; for the writing application gap, training task categories include sentence construction training, sentence transformation training, and writing application training; for the listening input gap, training task categories include listening comprehension training, keyword location training, and sound-meaning matching training; for the comprehensive application gap, training task categories include combined listening, speaking, reading, and writing training. Subsequently, the learning strategy generation module determines the weights of the training tasks based on the ability gap values. (The last sentence, "Let knowledge points...", is a separate, unrelated statement and doesn't need a direct translation.) The corresponding training task weights are The training task weights are then expressed as:

[0171] ;

[0172] in, Representing knowledge points The corresponding training task weights; Representing knowledge points The corresponding capability fault value; This represents the number of knowledge points involved in the ranking. The training task weights are used to characterize the allocation ratio of compensation training resources for different knowledge points.

[0173] After obtaining the training task categories and weights, a set of compensation training requirements is constructed. (The remaining text appears to be a fragmented list of knowledge points.) The corresponding set of training task categories is:

[0174] ;

[0175] in, Represents the set of training task categories; Indicates the first One training task category; This indicates the number of training task categories. The learning strategy generation module associates the training task categories with the training task weights to form a set of compensation training requirements, and generates a compensation training strategy based on this set. The compensation training strategy includes at least the training task category, training order, and training iteration allocation information.

[0176] After generating the compensation training strategy, the learning strategy generation module further acquires information on the importance of knowledge points and their learning status. The importance information represents the relevance of a knowledge point within the current curriculum, while the learning status information represents the current level of mastery of that knowledge point. Let the knowledge point... The corresponding importance value The learning state value is The learning priority is then expressed as:

[0177] ;

[0178] in, Representing knowledge points Corresponding learning priorities; Indicates the weights of the training task; This indicates the importance value of a knowledge point; This indicates the learning status value of a knowledge point. The learning strategy generation module sorts multiple knowledge points according to learning priority and sends the sorting results to the education service module for subsequent personalized learning path updates.

[0179] The above method can determine the corresponding training task category based on the type of capability gap, construct a set of compensation training requirements by combining capability gap values, and generate learning priorities by using the importance information of knowledge points and the learning status information of knowledge points. This enables the configuration of the association between training content and the capability difference status of knowledge points, and provides a basis for updating personalized learning paths.

[0180] Furthermore, the education service module connects with the learning strategy generation module to update personalized learning paths based on learning priorities and output corresponding compensatory training content.

[0181] Specifically, the education service module receives learning priorities, compensatory training strategies, and training task category information from the learning strategy generation module, and reorders multiple knowledge points in the current learning path according to the learning priorities. The education service module pre-establishes a knowledge point learning path library, which records the dependencies between knowledge points and the correspondence between course stages. For knowledge points... The education service module obtains the corresponding learning priority. The knowledge points are then sorted according to their learning priority. Let the sorted sequence of knowledge points be represented as:

[0182] ;

[0183] in, This indicates the updated personalized learning path; This indicates the knowledge point with the highest priority after sorting. This represents the knowledge point with the lowest priority after sorting. This indicates the number of knowledge points in the learning path. The education service module adjusts the position of knowledge points in the learning path based on the sorting results, while preserving the pre-defined dependencies between knowledge points, so that the updated learning path results meet the constraints of the course structure.

[0184] After updating the learning path, the education service module determines the corresponding compensatory training content based on the compensatory training strategy. Specifically, for knowledge points corresponding to gaps in oral output, it outputs scenario-based question-and-answer training, oral retelling training, and impromptu expression training; for knowledge points corresponding to gaps in writing application, it outputs sentence construction training, sentence transformation training, and writing application training; for knowledge points corresponding to gaps in listening input, it outputs listening comprehension training, keyword location training, and sound-meaning matching training; and for knowledge points corresponding to gaps in comprehensive application, it outputs combined listening, speaking, reading, and writing training. The education service module associates the training content with the corresponding knowledge points and generates a sequence of learning tasks according to the order in the learning path.

[0185] Set knowledge points The corresponding set of compensated training content is represented as follows:

[0186] ;

[0187] in, Representing knowledge points The corresponding set of compensation training content; Indicates the first One set of compensatory training content; This indicates the amount of supplementary training content. The education service module generates learning task outputs based on the updated personalized learning path and the set of supplementary training content, and sends them to the learning terminal for display.

[0188] After the learning terminal completes the corresponding training task, the education service module receives new learning data and sends it to the multimodal learning data acquisition module to form a new analysis cycle, thereby realizing the continuous updating of the learning path and compensation training content.

[0189] The above method enables the sorting and updating of knowledge points according to learning priorities, and the association output of the training content corresponding to the compensation training strategy with the knowledge points, so that the learning path update results and the compensation training content maintain a correspondence, thereby providing data output results for the execution of subsequent learning tasks.

[0190] Example 2: Refer to Figure 6 In a second embodiment of the present invention, the present invention provides a personalized English education method based on AI multimodal and big data, comprising the following steps:

[0191] S1. Collect learners' learning data in reading comprehension tasks, listening recognition tasks, vocabulary selection tasks, shadowing tasks, oral expression tasks, and writing application tasks, and generate multimodal learning data.

[0192] S2. Identify the corresponding knowledge points based on the multimodal learning data, establish the knowledge-task mapping relationship between the same knowledge point and multiple task types, and generate task performance data and task performance vectors.

[0193] S3. Construct cross-task performance distribution based on the task performance vector corresponding to the same knowledge point, calculate the cross-task consistency index, and generate knowledge point consistency analysis results.

[0194] S4. Calculate the performance values ​​of input tasks and output tasks based on the results of the knowledge point consistency analysis, and generate capability stratification values ​​based on the differences between the performance values ​​of input tasks and output tasks.

[0195] S5. Determine the type of capability gap based on the differences in task performance, the combination of tasks with the greatest differences, and the capability gap value.

[0196] S6. Determine the training task category based on the capability gap type, determine the training task weight based on the capability gap value, construct a set of compensation training requirements, and generate a compensation training strategy.

[0197] S7. Generate learning priorities based on the compensation training strategy, the importance of knowledge points, and the learning status of knowledge points; update the personalized learning path; and output the corresponding compensation training content.

[0198] Specifically, in step S1, the system collects learners' learning records in reading comprehension tasks, listening recognition tasks, vocabulary selection tasks, shadowing tasks, oral expression tasks, and writing application tasks, and performs correlation processing based on learner identifiers, knowledge point identifiers, and task type identifiers to generate corresponding multimodal learning data.

[0199] In step S2, the system parses words, phrases, sentence patterns, and semantic units in the multimodal learning data, identifies the corresponding knowledge points, and establishes a knowledge-task mapping relationship between the same knowledge point and multiple task types. Subsequently, the system acquires the task completion result data, task completion time data, task error count data, and task quality score data corresponding to the knowledge points, and generates task performance data and task performance vectors.

[0200] In step S3, the system constructs a cross-task performance distribution based on the task performance vector and calculates the corresponding cross-task consistency index to generate knowledge point consistency analysis results, so as to characterize the degree of consistency of the performance of the same knowledge point in different task types.

[0201] In step S4, the system divides the reading comprehension task, listening recognition task, and vocabulary selection task into input tasks, and the reading aloud task, oral expression task, and writing application task into output tasks. It then calculates the performance values ​​of the input tasks and the output tasks respectively, and generates a capability stratification value based on the difference between the two.

[0202] In step S5, the system calculates the performance differences between reading comprehension and oral expression tasks, between vocabulary selection and writing application tasks, and between listening recognition and shadowing tasks, determines the task combination with the greatest difference, and determines the type of ability gap by combining the ability gap value.

[0203] In step S6, the system determines the corresponding training task category based on the capability fault type, determines the training task weight based on the capability fault value, and constructs a set of compensation training requirements based on the training task category and the training task weight to generate a compensation training strategy.

[0204] In step S7, the system generates learning priorities based on the compensation training strategy, the importance of knowledge points, and the learning status of knowledge points. It then sorts multiple knowledge points, updates the personalized learning path, and outputs the corresponding compensation training content. New learning data generated after the learner completes compensation training re-enters step S1, thus achieving continuous updating of the personalized learning path.

[0205] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A personalized English education system based on AI multimodal and big data, characterized in that: include: The multimodal learning data acquisition module collects learners' learning data in reading comprehension tasks, listening recognition tasks, vocabulary selection tasks, shadowing tasks, oral expression tasks, and writing application tasks, and generates corresponding multimodal learning data. The knowledge analysis module is connected to the multimodal learning data acquisition module. Based on the multimodal learning data, it identifies the corresponding knowledge points, establishes a knowledge-task mapping relationship between the same knowledge point and multiple task types, and generates task performance data and task performance vectors. The cross-task consistency analysis module is connected to the knowledge analysis module. It constructs a cross-task performance distribution based on the task performance vector corresponding to the same knowledge point, calculates the corresponding cross-task consistency index, and generates knowledge point consistency analysis results. The capability gap identification module is connected to the cross-task consistency analysis module. It generates capability gap values ​​based on the knowledge point consistency analysis results and determines the capability gap type based on the capability gap values ​​and differences in task performance. The learning strategy generation module is connected to the capability gap identification module. It determines the training task category based on the capability gap type, determines the training task weight based on the capability gap value, constructs a set of compensation training requirements, generates a compensation training strategy based on the set of compensation training requirements, and generates a learning priority based on the compensation training strategy, knowledge point importance information, and knowledge point learning status information. The education service module is connected to the learning strategy generation module, updates the personalized learning path according to the learning priority, and outputs the corresponding compensation training content.

2. The personalized English education system based on AI multimodal and big data according to claim 1, characterized in that, In the knowledge analysis module, the step of establishing the knowledge-task mapping relationship between the same knowledge point and multiple task types includes: Extract knowledge point information from multimodal learning data; Identify the task type corresponding to each knowledge point; Establish a mapping relationship between the same knowledge point and multiple task types; A knowledge task mapping relationship is generated based on the aforementioned association mapping relationship.

3. The personalized English education system based on AI multimodal and big data according to claim 1, characterized in that, In the knowledge analysis module, the steps of generating task performance data and task performance vectors include: Obtain task completion result data, task completion time data, task error count data, and task quality score data for each knowledge point; Standardize the task completion result data, task completion time data, task error count data, and task quality score data. Generate task performance data based on the standardized processing results; Construct a task performance vector based on the performance data of multiple tasks corresponding to the same knowledge point.

4. The personalized English education system based on AI multimodal and big data according to claim 1, characterized in that, In the cross-task consistency analysis module, the step of calculating the corresponding cross-task consistency index includes: Obtain the task performance vector corresponding to the same knowledge point; Construct a cross-task performance distribution based on the task performance vector; Calculate the average performance value corresponding to the cross-task performance distribution; Calculate the degree of dispersion between the performance data of each task and the average performance value; A cross-task consistency index is generated based on the degree of dispersion. Knowledge point consistency analysis results are generated based on the cross-task consistency index.

5. The personalized English education system based on AI multimodal and big data according to claim 1, characterized in that, In the capability gap identification module, the step of generating capability gap values ​​based on the knowledge point consistency analysis results includes: Reading comprehension, listening recognition, and vocabulary selection tasks are categorized as input-type tasks. The tasks of reading aloud, oral expression, and writing application are classified as output tasks. Calculate the performance value of the input class task corresponding to the input class task; Calculate the performance value of the output class task corresponding to the output class task; Ability fault values ​​are generated based on the differences between the performance values ​​of input tasks and the performance values ​​of output tasks.

6. The personalized English education system based on AI multimodal and big data according to claim 1, characterized in that, In the capability fault identification module, the step of determining the type of capability fault includes: Calculate the performance differences between reading comprehension and oral expression tasks; Compute the performance differences between vocabulary selection tasks and writing application tasks; Calculate the performance differences between the listening recognition task and the reading aloud task; Identify the combination of tasks with the greatest performance differences; The capability fault type is determined based on the maximum difference task combination and the capability fault value; The types of ability gaps include oral output gaps, writing application gaps, listening input gaps, and comprehensive application gaps.

7. The personalized English education system based on AI multimodal and big data according to claim 1, characterized in that, In the learning strategy generation module, the step of generating a compensation training strategy based on the compensation training requirement set includes: Identify the capability gap type; determine the training task category based on the capability gap type; The weights of the training tasks are determined based on the capability gap values; Construct a set of compensation training requirements based on training task categories and training task weights; A compensation training strategy is generated based on the set of compensation training requirements.

8. The personalized English education system based on AI multimodal and big data according to claim 1, characterized in that, In the learning strategy generation module, the step of generating learning priorities includes: Acquire compensatory training strategies; acquire information on the importance of knowledge points; Obtain information on the learning status of knowledge points; Learning priorities are generated based on the compensation training strategy, the importance of knowledge points, and the learning status of knowledge points; multiple knowledge points are then sorted according to the learning priorities.

9. The personalized English education system based on AI multimodal and big data according to claim 7, characterized in that, The step of determining the training task category based on the capability gap type includes: When the type of ability gap is oral output gap, determine the situational question and answer training, oral retelling training and impromptu expression training; When the type of competency gap is writing application gap, the training should include sentence construction training, sentence transformation training, and writing application training. When the ability gap type is auditory input gap, determine auditory discrimination training, keyword localization training, and sound-meaning matching training; When the competency gap type is a comprehensive application gap, determine the combined training of listening, speaking, reading and writing.

10. A personalized English education method based on AI multimodal and big data, characterized in that: The personalized English education system based on AI multimodal and big data, as described in any one of claims 1-9, includes the following steps: S1. Collect learners' learning data in reading comprehension tasks, listening recognition tasks, vocabulary selection tasks, shadowing tasks, oral expression tasks, and writing application tasks, and generate multimodal learning data. S2. Identify the corresponding knowledge points based on the multimodal learning data, establish a knowledge-task mapping relationship between the same knowledge point and multiple task types, and generate task performance data and task performance vectors. S3. Construct cross-task performance distribution based on the task performance vector corresponding to the same knowledge point, calculate the cross-task consistency index, and generate knowledge point consistency analysis results. S4. Calculate the performance values ​​of input tasks and output tasks based on the consistency analysis results of the knowledge points, and generate capability stratification values ​​based on the differences between the performance values ​​of input tasks and output tasks. S5. Determine the type of capability gap based on the differences in task performance, the combination of tasks with the greatest differences, and the capability gap value. S6. Determine the training task category based on the capability fault type, determine the training task weight based on the capability fault value, construct a set of compensation training requirements, and generate a compensation training strategy. S7. Generate learning priorities based on the compensation training strategy, knowledge point importance information, and knowledge point learning status information, update personalized learning paths, and output corresponding compensation training content.

Images