Adaptive Teaching System and Method for Bilingual Story Comprehension Based on Individual Differences

Abstract

This invention relates to the field of bilingual teaching technology, specifically to an adaptive bilingual story comprehension teaching system and method tailored to individual differences. The system includes a structural adaptation analysis module, a comprehension feature summarization module, a material matching and evaluation module, an expression trend tracking module, and a strategy configuration adjustment module. This invention, through dynamic identification based on multi-dimensional features such as roles, sequence, syntax, keywords, semantics, and visual perception, comprehensively analyzes students' unique differences in story comprehension, expression content, language transfer, and cognitive responses. It continuously tracks material content, comprehension processes, and expression trajectories, automatically generating individual characteristic parameters. This enables refined selection of story fragments and adjustment of task difficulty, real-time summarization of expression evolution trends, and dynamic adjustment of intervention configurations. This promotes the continuous improvement of language structure application, cognitive transfer, and diverse expression abilities, fosters a high degree of adaptation between teaching content and student abilities, and enhances the personalization of learning paths and the timeliness of feedback regulation.

Description

Technical Field

[0001] This invention relates to the field of bilingual teaching technology, and in particular to an adaptive bilingual story comprehension teaching system and method that addresses individual differences. Background Technology

[0002] The field of bilingual education involves using two or more languages ​​as a medium to promote learners' comprehensive improvement in language proficiency and intercultural communication skills through diverse language activities such as listening, speaking, reading, and writing. This includes instructional content design based on the inherent laws of language, multimodal language input and output, analysis of language cognitive processes, and teaching evaluation methods. Overall, it encompasses interdisciplinary content from cognitive science, linguistics, and educational informatization, emphasizing the optimization of learning paths through scientific methods and data analysis to adapt to the needs of different learners. Among these, traditional bilingual story comprehension adaptive teaching systems address the differences in learners' language abilities during story comprehension. This typically involves differentiated instruction based on teacher experience, textbook grading, and post-class quizzes, adjusting classroom difficulty and pace according to a pre-set teaching process and content. This is generally accomplished through manual grading, manual assessment, and static selection of teaching content.

[0003] Existing technologies largely rely on manual stratification and textbook grading, employing static content and post-class quizzes. They lack the ability to track the multidimensional expressive characteristics and stage changes during the learning process, making it difficult to dynamically monitor learners' actual language development. In real classrooms with diverse contexts and constantly changing expressions, there is often a mismatch between teaching content and students' abilities, resulting in delayed teaching adjustments, limited feedback, and some students remaining at unsuitable teaching levels for extended periods, leading to comprehension biases, decreased interest, and slow skill improvement. Summary of the Invention

[0004] The purpose of this invention is to address the shortcomings of existing technologies by proposing an adaptive bilingual story comprehension teaching system and method that addresses individual differences.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: a bilingual story comprehension adaptive teaching system oriented towards individual differences, the system comprising:

[0006] The structural adaptation analysis module is based on role-switching bilingual texts. It analyzes the number of roles, the order of events and the content of students' retelling, determines the actual distribution of temporal adverbs and titles of people, compares the order of the retelling with the original text, and combines the perception level classification data to summarize the differences in language structure application and obtain structural adaptation performance parameters.

[0007] Based on the structural adaptation performance parameters, the understanding feature induction module calculates the number of keyword responses in bilingual plot question-and-answer, analyzes the syntactic structure integrity, compares the semantic preservation during bilingual translation, and, in conjunction with type criteria and transfer strength, summarizes language understanding changes to obtain the understanding feature offset.

[0008] Based on the aforementioned understanding feature offset, the material matching evaluation module filters out segments with structural differences in the material library, compares the performance of sentence pattern recognition with the lag in visual understanding, determines the distribution of semantic connection interruptions, analyzes the differences in story structure, and obtains the structural matching distribution index.

[0009] Based on the structure matching distribution index, the expression trend tracking module optimizes the segment expression in bilingual story retelling detection, counts the frequency of word recurrence and the number of sentence pattern transformations, analyzes the changes in sentence group length and density, and obtains the expression trend trajectory group by combining expression normation parameters.

[0010] The present invention is improved in that the structural adaptation performance parameters include expression accuracy, language structure complexity, and information integration level; the understanding feature offset includes semantic extraction sensitivity, knowledge transfer amplitude, and understanding stability; the structural matching distribution index includes material diversity distribution, structural fit, and content association strength; and the expression trend trajectory group includes expression pattern changes, content presentation trends, and repetition coherence.

[0011] The present invention is improved in that the structural adaptation analysis module includes:

[0012] The role distribution extraction submodule extracts the number of roles and their first appearance order in the text based on the role-converting bilingual text. Combined with the student's retelling text, it identifies the names of the characters one by one. By comparing the appearance order of the character names in the original text and the retelling text, it counts the frequency of the misalignment of the first appearance order of the characters between the two texts and obtains the difference in the character order.

[0013] The temporal distribution comparison submodule calls the role order difference quantity to detect all temporal adverbs in the student's retelling text, distinguishes their distribution range and frequency of use by paragraph, arranges the adverbs used for each event sequence in the original text to correspond, and compares the paragraph coverage and mutual exclusion distribution of time markers in the two texts to obtain the event sequence misalignment rate.

[0014] The structural adaptation induction submodule analyzes the frequency of conjunctions, syntactic nesting levels, subject substitution frequency, and sentence group transition density in the students' retelling content based on the event sequence misalignment rate. It also combines the plot perception level classification data and performs type interval matching on the features to obtain structural adaptation performance parameters.

[0015] The present invention is improved in that the understanding feature induction module includes:

[0016] Based on the aforementioned structural adaptation performance parameters, the keyword response submodule statistically analyzes the responses of students to all plot keywords in the bilingual plot question-and-answer task, distinguishes the different task stages, categorizes the position and repetition frequency of keywords in each stage, and accumulates and sums the number of keyword responses to obtain the total number of keyword responses.

[0017] The syntactic structure detection submodule decomposes the sentences of the response content based on the total number of keyword responses, retrieves the completeness of the subject, predicate, object and modifiers, and classifies the syntactic structure coverage performance according to the sentence construction type standard to obtain the syntactic structure integrity level coefficient.

[0018] The semantic transfer comparison submodule extracts bilingual translation content based on the syntactic structure integrity level coefficient, decomposes the target language vocabulary and semantic information, compares the source language information retention performance, and obtains the understanding feature offset by combining the transfer strength and sentence change index.

[0019] The present invention is improved in that the material matching and evaluation module includes:

[0020] Based on the understanding feature offset, the segment filtering submodule obtains the segment structure features of the combined text and image story material library, collects the number of sentence patterns, the total number of information levels and the frequency of character perspective changes for each segment, analyzes the structural span based on the differences in sentence patterns, information levels and perspective frequency, and obtains a set of structural difference segments by filtering segments that meet the conditions.

[0021] The performance recognition submodule collects students' recognition records of segments in bilingual listening and reading tasks based on the set of structural difference segments, extracts the correct frequency of sentence pattern recognition, task response time and error expression type code for each segment, compares the recognition frequency with the reference standard data, analyzes the impact of the combination of response time and error expression on the rhythm of understanding, and obtains the sentence pattern comprehension lag set.

[0022] The structural comparison submodule, based on the lag set of the sentence pattern comprehension, compares the visual pause distribution, total number of narrative jumps, and comprehension delay performance of each segment to obtain the degree of matching deviation, analyzes the degree of correspondence with the evolution path of the original structural style of the segment, and obtains the structural matching distribution index.

[0023] The present invention is improved in that the expression trend tracking module includes:

[0024] The vocabulary recurrence submodule extracts keywords from the detected content based on the structure matching distribution index, counts the frequency of occurrence of words in each detection task segment, distinguishes the distribution of words in segments with different time sequences, and obtains the vocabulary recurrence density level based on the difference between the number of repetitions and word class coverage within the segment.

[0025] Based on the vocabulary recurrence density level, the sentence pattern transformation submodule counts the number of sentence pattern types in each detected segment, distinguishes the types of changes in sentence pattern structure between adjacent segments, compares the usage of different sentence pattern combinations, summarizes the types of sentence pattern differences and their usage frequency, and obtains the differences in sentence pattern transformation distribution.

[0026] The expression ratio analysis submodule extracts the sentence group length and sentence type number of each segment in the detected content based on the differences in sentence pattern transformation distribution. It compares each structure according to the expression standard parameters, calculates the expression density offset of each segment, integrates the related expression density data, and obtains the expression trend trajectory group.

[0027] The present invention has an improvement, wherein the system further includes:

[0028] Based on the expression trend trajectory group, the strategy configuration adjustment module adjusts the task arrangement rhythm, judges the semantic prompt interval setting, analyzes the matching between text and image intervention and expression trajectory, compares the intervention configuration positioning, summarizes parameter changes, and obtains the intervention parameter offset magnitude.

[0029] The deviation range of the intervention parameters includes the adjustment range of task distribution, the sensitivity of intervention response, and the level of strategy matching.

[0030] The present invention is improved in that the strategy configuration adjustment module includes:

[0031] The prompt rhythm judgment submodule analyzes the arrangement rhythm of each task in the teaching library based on the expression trend trajectory group, compares it with the difference in the distribution of graphic intervention prompts, judges the density change of semantic intervention prompts in each task interval, and filters out the abnormal intervals with the best correlation with the turning point of the paragraph to obtain the prompt interval offset.

[0032] Based on the prompt interval offset, the intervention fusion evaluation submodule analyzes the distribution of the text-image intervention fusion density on the segment expression nodes, determines the correspondence between the intervention configuration and the node type in each time period of the expression, optimizes the intervention distribution order, and obtains the intervention configuration positioning frequency.

[0033] The configuration offset calculation submodule analyzes the differences in the density of text-image intervention fusion, sentence density ratio, vocabulary recurrence frequency, and paragraph expression type based on the frequency of the intervention configuration positioning, compares the changes in structural arrangement and expression mode, and obtains the magnitude of the intervention parameter offset.

[0034] An adaptive teaching method for bilingual story comprehension based on individual differences, which is implemented based on the aforementioned adaptive teaching system for bilingual story comprehension based on individual differences, includes the following steps:

[0035] S1: Based on role-switching bilingual texts, analyze the number of characters and the order of events. Combined with the content of students' bilingual story retelling, determine the actual distribution of temporal adverbs and character titles used by students in the retelling text. Compare the correspondence between the order of students' retelling and the order of the original text. Based on the data of the level of perception of the story plot, obtain the structural adaptation performance parameters.

[0036] S2: Based on the structural adaptation performance parameters, calculate the number of times students respond to keywords involved in the bilingual scenario question and answer process, analyze the complete performance of syntactic structure in the answer content, compare the semantic preservation in the bilingual translation process, summarize the changes in students' language comprehension characteristics, and obtain the comprehension feature offset.

[0037] S3: Based on the aforementioned understanding feature offset, screen story fragments with significant structural differences within the combined text and image story material library, compare students' performance in sentence pattern recognition and visual comprehension lag in bilingual listening and reading tasks, determine the distribution characteristics of semantic connection breakpoints, and analyze the differences between the story fragment structure and students' comprehension performance to obtain structural matching distribution indicators.

[0038] S4: Based on the structure matching distribution index, optimize the expression of segments in the bilingual story retelling test process, count the frequency of word recurrence in the test content, measure the number of sentence pattern transformations, analyze the ratio change of sentence group length and sentence pattern density, summarize the distribution trend of expression mode, and obtain the expression trend trajectory group.

[0039] S5: Based on the expression trend trajectory group, adjust the task arrangement rhythm in the teaching library, determine the setting of semantic intervention prompts in each interval, analyze the matching relationship between the density of graphic intervention fusion and the student's expression trajectory, compare the positioning of intervention configuration in the expression process, and obtain the magnitude of intervention parameter offset.

[0040] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0041] In this invention, dynamic recognition based on multi-dimensional features such as role, time sequence, syntax, keywords, semantics, and vision is used to comprehensively analyze the unique differences among students in story comprehension, expression, language transfer, and cognitive response. By linking the continuous tracking of material content, comprehension process, and expression trajectory, individual feature parameters are automatically generated. This enables refined selection of story fragments and adjustment of task difficulty, real-time summarization of expression evolution trends, dynamic adjustment of intervention configuration, and continuous improvement of language structure application, cognitive transfer, and diverse expression abilities. It also promotes a high degree of fit between teaching content and students' abilities, enhances the personalization of learning paths, and improves the timeliness of feedback regulation. Attached Figure Description

[0042] Figure 1 This is a system flowchart of the present invention;

[0043] Figure 2 This is a flowchart of the structural adaptation analysis module in this invention;

[0044] Figure 3 This is a flowchart of the feature induction module in this invention;

[0045] Figure 4 This is a flowchart of the material matching and evaluation module in this invention;

[0046] Figure 5 This is a flowchart illustrating the trend tracking module in this invention;

[0047] Figure 6 This is a flowchart of the strategy configuration adjustment module in this invention. Detailed Implementation

[0048] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0049] In the description of this invention, it should be understood that the terms "length," "width," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings and are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, in the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0050] Example 1, please refer to Figure 1 This invention provides a technical solution: a bilingual story comprehension adaptive teaching system for individual differences, comprising:

[0051] The structural adaptation analysis module is based on role-switching bilingual texts. It analyzes the number of characters and the order of events, combines the content of students' bilingual story retelling, determines the actual distribution of temporal adverbs and character titles used by students in the retelling text, compares the correspondence between the order of students' retelling and the order of the original text, and summarizes students' adaptation performance in language structure application based on the tiered data of story perception level, thus obtaining structural adaptation performance parameters.

[0052] The understanding feature summarization module, based on structural adaptation performance parameters, calculates the number of times students respond to keywords involved in bilingual scenario-based question-and-answer sessions, analyzes the completeness of syntactic structure in the answers, compares the semantic preservation in the bilingual translation process, and, in conjunction with sentence construction type standards, keyword extraction sets, and language transfer strength references, summarizes changes in students' language comprehension features to obtain comprehension feature offsets.

[0053] The material matching evaluation module, based on the understanding feature offset, filters story fragments with significant structural differences within the combined text and image story material library. It compares students' performance in sentence pattern recognition and visual comprehension lag in bilingual listening and reading tasks, determines the distribution characteristics of semantic connection breakpoints, and analyzes the differences between the story fragment structure and students' comprehension performance to obtain structural matching distribution indicators.

[0054] The expression trend tracking module optimizes the expression of segments in the bilingual story retelling test process based on the structure matching distribution index, counts the frequency of word recurrence in the test content, measures the number of sentence pattern transformations, analyzes the ratio change of sentence group length and sentence pattern density, and summarizes the expression mode distribution trend by combining the corresponding parameters of expression norms, thus obtaining the expression trend trajectory group.

[0055] The strategy configuration adjustment module adjusts the task arrangement rhythm in the teaching library based on the expression trend trajectory group, judges the setting of semantic intervention prompts in each interval, analyzes the matching relationship between the density of graphic intervention integration and the student's expression trajectory, compares the positioning of intervention configuration in the expression process, summarizes the changes in intervention parameter configuration, and obtains the magnitude of intervention parameter offset.

[0056] Structural adaptation performance parameters include expression accuracy, language structure complexity, and information integration level; comprehension feature offset includes semantic extraction sensitivity, knowledge transfer magnitude, and comprehension stability; structural matching distribution indicators include material diversity distribution, structural fit, and content relevance strength; expression trend trajectory group includes expression pattern changes, content presentation trends, and repetition coherence; and intervention parameter offset magnitude includes task distribution adjustment magnitude, intervention response sensitivity, and strategy matching level.

[0057] In Module 1, role-switching bilingual texts refer to bilingual (e.g., Chinese and English) story materials containing multiple characters and involving shifts in character identity, language expression, or perspective during the narrative process; temporal adverbs refer to adverbs used in story narration or retelling to express the order of time and the sequence of events (e.g., "first," "then," "finally," "afterwards," "meanwhile," etc.); actual distribution in retelling texts refers to the position, frequency, and distribution pattern of temporal adverbs or character titles in students' narration when retelling the story; correspondence refers to the specific one-to-one matching or inconsistency between students' retelling content and the original story content in terms of character order, event order, etc.; perceptual level grading data refers to data parameters formed by dividing students' ability to understand and express the plot into several levels or intervals based on educational psychology, curriculum standards, etc.; adaptive performance refers to the level of adaptation and performance of students in terms of story structure, language expression, and role switching during the retelling or understanding process.

[0058] In Module 2, the frequency of keyword responses refers to the statistical count of the number of times students accurately identify, mention, or use target keywords (such as characters, actions, plot points, etc.) in bilingual scenario-based question-and-answer or tests; completeness refers to the degree of completeness of sentences or paragraphs in syntactic structure (such as subject-verb-object completeness, complete modifiers, etc.) in tasks such as answering and retelling; semantic retention refers to the degree to which the original meaning, information content, and logical relationships are completely preserved or lost when students translate information from one language to another; construction type standard refers to the authoritative standard or classification table used in the field of language education to distinguish sentence structure types (such as simple sentences, compound sentences, complex sentences, etc.) or expression patterns; language transfer intensity reference refers to the reference standard or quantitative range for evaluating the degree to which students' mother tongue knowledge affects their second language expression during the bilingual conversion process; and language comprehension characteristic changes refer to the changes in students' language comprehension ability, methods, and accuracy under different tasks, different stages, and different materials.

[0059] In Module 3, "distinct story fragments" refers to bilingual story fragments that differ significantly from students' current abilities in terms of sentence structure, content, and difficulty of expression; "sentence structure recognition performance" refers to the types and number of sentence structures that students can accurately identify and understand in listening or reading tasks, as well as their actual performance; "visual comprehension lag" refers to situations where students' reaction speed or synchronicity in understanding image and text information is insufficient when receiving illustrated materials; and "interruption point distribution characteristics" refers to the specific locations and distribution patterns of interruptions or pauses in the semantic flow (i.e., comprehension obstacles, jumps, missing information, etc.) during the process of understanding the material.

[0060] In Module 4, paragraph expression refers to the students' ability to retell or express a complete paragraph of a story as required during the assessment or test. Vocabulary recurrence frequency refers to the frequency with which the same word is used multiple times in a student's retelling or expression, reflecting the richness or uniformity of their vocabulary application. Proportional changes refer to the analysis of the proportions and trends between indicators such as sentence group length and sentence density in different assessment content or task stages. Standardized parameters related to expression ability refer to standardized parameters such as the usage, sentence patterns, or structural norms specified in textbooks, curriculum standards, and level examinations. The distribution trend of expression methods refers to the frequency and temporal variation patterns of various expression methods (such as simple description, complex retelling, and inferential expression) during the students' expression process.

[0061] In Module 5, the task arrangement rhythm refers to the sequence, switching speed, and rhythm of task segments in teaching activities, i.e., the distribution and progress rhythm of each learning task; the setting within each interval refers to the division and specific arrangement of intervals with different teaching parameters or prompts in different tasks, segments, and difficulty levels; the matching relationship refers to the degree of fit and correspondence between teaching intervention measures (such as text-image combination, task allocation, semantic prompts, etc.) and students' actual learning process and performance; the positioning in the expression process refers to the specific location of which segment and stage different intervention configurations are applied in students' expression activities; and the change in intervention parameter configuration refers to the adjustment and change of intervention methods or parameter settings used in teaching as students' performance or task needs differ.

[0062] Please see Figure 2 The structural adaptation analysis module includes:

[0063] The role distribution extraction submodule extracts the number of roles and their first appearance order in the text based on the role-converting bilingual text. Combined with the student's retelling text, it identifies the names of the characters one by one. By comparing the appearance order of the character names in the original text and the retelling text, it counts the frequency of the misalignment of the first appearance order of the characters between the two texts and obtains the difference in the character order.

[0064] For bilingual texts involving role-switching, all character words with personal pronouns need to be extracted. First, the original bilingual text is read, and the names or pronoun forms of each character are selected, such as "Mom," "Tom," "the boy," "he," and "Xiaoming" in the Chinese and English stories. These are then numbered according to their first appearance in the text, forming a character sequence. Next, the student's retelling is read, and all character names are identified and extracted. It is then determined whether the student used the same character names or alternative expressions as the original text, such as replacing "Xiaohong" with "the girl." The first appearance of each character name in the student's retelling is compared with the original text. The order of characters in the text is compared one by one. When the order in which a character appears in the retelling is different from the order in which they first appear in the original text, it is recorded as a misalignment. For example, if the characters in the original text are A, B, C, and D, and the order in the student's retelling is B, A, D, and C, then the positions of the four characters A, B, C, and D do not correspond to the original text, and the misalignment frequency is 4. By comparing the order of the first appearance of all characters in the two texts one by one, the cumulative number of character misalignments is summarized as the character order difference. Assuming that the number of times the character order in the student's retelling matches the original text is 3 and the number of times it does not match is 2, then the character order difference is 2. This difference value will be used in the downstream module as a reference benchmark for analyzing the student's mastery of the story structure.

[0065] The temporal distribution comparison submodule calls the role order difference quantity, detects all temporal adverbs in the student's retelling text, distinguishes their distribution range and frequency of use by paragraph, arranges the adverbs used for each event sequence in the original text to correspond, and compares the paragraph coverage and mutual exclusion distribution of time markers in the two texts to obtain the event order misalignment rate.

[0066] All time-sequence adverbs in the student's retelling of the text were extracted and statistically analyzed word by word, including phrases such as "first," "then," "finally," "afterwards," and "meanwhile." Their distribution and frequency were recorded according to the paragraph in which they appeared, forming an adverb distribution table. The original text's time-sequence adverbs were processed in the same way according to the order of events, forming an original timeline adverb distribution table. The adverbs of the same type in the two tables were compared paragraph by paragraph. For example, if the original text uses "first" to describe event A in the first paragraph and "then" to describe event B in the second paragraph, and the student uses "then" in the first paragraph and "first" in the second paragraph, then the adverbs in the two paragraphs will be compared. When word distributions are mutually exclusive, it is recorded as an event sequence misalignment. This type of adverb comparison is performed on all paragraphs in the entire text. When the temporal adverb used in a paraphrased paragraph does not match the temporal order of the original paragraphs, that paragraph is recorded as a misaligned paragraph. The total number of misaligned paragraphs is counted and compared with the total number of paragraphs to obtain the event sequence misalignment rate. For example, if the original text has 5 paragraphs and 2 paragraphs in the student's paraphrase have misaligned temporal adverbs, the event sequence misalignment rate is 0.4. According to the established standards, a misalignment rate between 0 and 0.2 is considered low misalignment, between 0.2 and 0.5 is considered medium misalignment, and greater than 0.5 is considered high misalignment. This misalignment rate will be used as a parameter to determine the accuracy of connections and transitions in downstream structural adaptation analysis.

[0067] The structural adaptation induction submodule analyzes the frequency of conjunctions, syntactic nesting levels, subject replacement frequency, and sentence group transition density in students' retelling content based on the event sequence misalignment rate. It also combines the plot perception level classification data and performs type interval matching on the features to obtain structural adaptation performance parameters.

[0068] The use of conjunctions in students' retelling was extracted, and the frequency of each type of conjunction, such as "because," "but," "therefore," "and," and "but," was counted. The total frequency was calculated, and the average conjunction usage density was divided according to the number of sentences in each paragraph. Next, the structure of each sentence was read, and its syntactic structure was analyzed to identify the nesting level, such as whether a sentence contains modifying clauses, object clauses, or adverbial elements. After recording the nesting level of each sentence, the average nesting depth was calculated. The use of subjects in the text was analyzed, and whether subject changes occurred in consecutive sentences, such as expressions like "I→Tom→he→theboy," were recorded to determine whether the subject was replaced. The frequency of subject replacement in the entire text was counted. The entire paraphrased text is divided into sentence groups, for example, every three consecutive sentences form a sentence group. The connection between each sentence group is counted, whether it is established by adverbs or conjunctions. The ratio between the number of sentence groups constructed by transitions and the total number of sentence groups is recorded. The values ​​are compared with existing perception level classification data. For example, the average sentence group length of standard students in the upper grades of primary school is set to 15 characters, the average density of conjunctions is 3 per paragraph, and the nesting level of clauses is 2. If a student has a conjunction density of 2.5 and a nesting level of 1.5 in paraphrasing, it is matched as a low adaptation interval. By classifying the four indicators of conjunction frequency, nesting level, subject replacement frequency, and sentence group transition density into the corresponding interval according to the level, the structural adaptation performance parameters such as expression accuracy, language structure complexity, and information integration level are summarized and output.

[0069] Please see Figure 3 The feature induction module includes:

[0070] The keyword response submodule, based on the structural adaptation performance parameters, counts the responses of students to all plot keywords in the bilingual plot question-and-answer task, distinguishes the different task stages, classifies the position and repetition of keywords in each stage, and accumulates and sums the number of keyword responses to obtain the total number of keyword responses.

[0071] Structural adaptation performance parameters are used as pre-inputs to assess students' keyword response behavior in bilingual scenario-based question-and-answer tasks. First, the text content of the task is structurally deconstructed to extract core scenario keywords from the question stem and standard answers. For example, words like "girl," "opened the door," "found a gift," and "felt surprised" are marked in the Chinese and English question-and-answer materials and used as a target keyword list. Next, student responses are retrieved and compared to the phased prompts or task steps in the task settings. For example, phase one asks about character identity, phase two asks about the motivation of the event, and phase three asks about emotional feedback. Keyword hits in student responses are extracted at each phase, and word matching is used for comparison during the extraction process to record the hit location and the frequency of the same keyword. For example, when a student answers "Who opened the door?", they reply "the girl opened the door." If "the door and the girl was very surprised" appears twice, "the girl" appears once, "opened the door" appears once, and "surprised" appears once. These are recorded in the keyword response record. The number of keywords in different task stages is categorized. For example, if 3 keywords are hit in stage 1, 2 in stage 2, and 1 in stage 3, the total number of keyword responses across all stages is 6. The frequency of repeated use is also included in the total. For example, if "the girl" is repeated 3 times, the total number of responses is 3, not 1. The sum of the number of keyword responses across all task stages is the total number of keyword responses. In this round of tasks, this total can be output as 12 times. Based on the response density, the low response range is 0-4, the medium response range is 5-9, and the high response range is 10 and above.

[0072] The syntactic structure detection submodule decomposes the sentences of the response content based on the total number of keyword responses, retrieves the completeness of the subject, predicate, object and modifiers, and judges the syntactic structure coverage performance according to the sentence construction type standard to obtain the syntactic structure completeness level coefficient.

[0073] Using the total number of keyword responses as input, the sentence structure used by students in the question-and-answer session is broken down sentence by sentence. First, the text of each student's answer is read one by one, and each sentence is analyzed to determine if it completely contains core syntactic components such as subject, predicate, and object. Simultaneously, it checks for the presence of modifying elements such as adjectives, adverbs, or prepositional phrases. For example, if a student answers "the boy quickly ran in other rooms," this sentence contains "the boy" as the subject, "ran" as the predicate, "in other rooms" as the object, and "quickly" as a modifier, and is therefore considered structurally complete. If the answer is "ran quickly," the subject is missing, and the sentence is considered structurally incomplete. This type of breakdown and retrieval operation is performed sentence by sentence across all answers, and each sentence is marked. The system assigns a completeness label to each sentence and ranks them according to their structure. Each structure label is then assigned a value based on the sentence construction type standard. For example, a complete subject-verb-object structure is grade 3, a subject-verb structure is grade 2, and a sentence with only a predicate or subject is grade 1. If a compound sentence or complex sentence appears, it is weighted according to the nesting level, reaching grade 4 or 5. After classifying all sentences into these structural grades, the average or mode of all grade values ​​is used to determine the student's syntactic structural completeness grade coefficient in this response. If 6 out of 10 sentences are subject-verb-object structures, 3 are subject-verb structures, and 1 lacks a subject, the grade weights are 3, 2, and 1 respectively, resulting in a total weighted value of 25 and an average grade of 2.5. Based on the defined range of 2.5 to 3.0, which is considered a medium-to-high structural completeness range, the student's syntactic structural completeness grade coefficient in this response is ultimately determined to be medium-to-high.

[0074] The semantic transfer comparison submodule extracts bilingual translation content based on the syntactic structure integrity level coefficient, decomposes the target language vocabulary and semantic information, compares the source language information retention performance, and combines transfer strength and sentence structure change indicators, using the following formula:

[0075]

[0076] We obtain the understood feature offset ΔLS, where Qs i Represents the semantic encoding of the i-th keyword in the source language, Qt i Qm represents the semantic encoding of the i-th corresponding translation unit in the target language. i Qc represents the content saliency weight of the i-th keyword. i Qr represents the component coverage of the syntactic structure in the i-th sentence. j represents the number of transfer semantic defects in the j-th translated sentence, k represents the total number of translated sentences, and n represents the total number of keywords.

[0077] The comprehension feature offset is used to characterize the degree of comprehensive shift in students' language comprehension features caused by factors such as language transfer, expression transformation, and structural changes in bilingual story comprehension and retelling tasks. The higher the value of this index, the weaker the semantic retention, the more incomplete the transfer, and the lower the structural matching degree in the process of language comprehension and expression. Conversely, it indicates that the language comprehension process is more stable, accurate, and the transfer is smoother.

[0078] The structure of the Chinese-English translation passages completed by students in bilingual tasks is extracted, and the keywords and semantic units are vectorized and encoded. A source language encoding set Qs is constructed based on the semantic reference of each keyword in the source language. i, Simultaneously, a target language encoding set Qt is constructed based on the corresponding expressions in the students' translated texts. i, Taking the keywords "river," "build," and "dangerous" used by sample students in the bilingual task with a "natural disaster scenario" as an example, their original semantic strength scores were 0.92, 0.85, and 0.89, respectively, while the semantic preservation scores in the translated sentences were 0.87, 0.75, and 0.83, respectively. The normalized codes after semantic domain standardization are as follows:

[0079]

[0080] Meanwhile, in this task, the keyword salience weights were set as follows: "river" as the main environmental element, Qm1=0.85; "build" as an action verb, Qm2=0.95; and "dangerous" as an emotional adjective, Qm3=0.75. Based on the results from the syntactic parsing module, the distribution of the main structural integrity in the students' corresponding sentences was 3, 4, and 3, with corresponding normalized structure coverage numbers of Qc1=0.88, Qc2=0.93, and Qc3=0.88. For information gaps in the translation results, the system recorded the number of transfer semantic gaps as Qr1=0.20, Qr2=0.10, and Qr3=0.30, k=3, n=3. Substituting these data into the formula, the first step is to calculate the absolute difference multiplied by the weights:

[0081]

[0082] Total = 0.0255 + 0.0665 + 0.0375 = 0.1295;

[0083] The second step is to calculate the square root of the denominator:

[0084]

[0085] Third step, calculating the first fraction:

[0086]

[0087] Step 4, Calculate the mean of the second item:

[0088]

[0089] Combined calculation results:

[0090]

[0091] The results indicate that when the calculated comprehension feature offset ΔLS = 0.2833 falls within the set intermediate offset interval [0.25, 0.40], it indicates that students exhibit moderate semantic preservation bias and structural expression instability during bilingual translation. Specifically, this manifests as partial semantic weakening in keyword translation, insufficient syntactic structural support, and a high frequency of information incompleteness in translated sentences. The formula, by introducing the cross-interaction mechanism between keyword semantic difference, significance weight, structural integrity factor, and transfer incompleteness rate, realizes a numerical description of cross-linguistic semantic alignment offset, providing a continuous and measurable variable input for subsequent automatic adjustment of material structure and strategy configuration.

[0092] Please see Figure 4 The material matching and evaluation module includes:

[0093] The segment filtering submodule, based on understanding feature offsets, obtains the segment structure features of the combined text and image story material library, collects the number of sentence patterns, the total number of information levels, and the frequency of character perspective changes for each segment, analyzes the structural span based on the differences in sentence patterns, information levels, and perspective frequency, and obtains a set of structural difference segments by filtering segments that meet the conditions.

[0094] The structural features of each segment are extracted one by one. First, the text is broken down into sentences, and the number of sentence patterns is counted. Sentence patterns such as declarative sentences, interrogative sentences, exclamatory sentences, passive sentences, and coordinate sentences are grouped and categorized according to their frequency of occurrence. The total number of each type of sentence pattern is calculated, and the frequency of each type is recorded. For example, if a segment contains four types of sentence patterns (10 declarative sentences, 3 passive sentences, 2 coordinate sentences, and 1 interrogative sentence), the total number of sentence patterns is 4, and the total number of sentence patterns is 16. Next, the information hierarchy of the text is evaluated. By comparing the paragraph structure with image annotations, various information points in the text are divided into first-level, second-level, and third-level information levels according to whether they are main events, detailed descriptions, background information, or emotional descriptions. The number of different information levels in the segment is counted. For example, if a segment contains 2 first-level events, 3 second-level details, and 1 third-level emotion, the total number of information levels is 6. Simultaneously, the frequency of changes in the perspectives of the characters involved in the segment is scanned. A change in perspective is considered a shift in narrative angle within a segment, whether it's from first-person to third-person or from a single character's description to a multi-character interweaving. The number of perspective shifts in each segment is counted. For example, if a segment contains three perspective transitions such as "he thought..." and "he actually...", the frequency of perspective shifts is 3. The differences among these three parameters are then compared. The maximum and minimum values ​​of sentence structure differences, information level spans, and perspective shift spans among all segments in the same batch of materials are calculated. Thresholds are then used to determine if the span exceeds the structural selection criteria. The sentence structure difference screening threshold is set at 3 categories, the information level span threshold at 4 levels, and the perspective shift frequency threshold at 2 times. If a segment exceeds any of these thresholds, its structural span is considered large, and it is selected into the structural difference segment set. For example, segment A has 5 sentence structure categories, 7 information levels, and 5 perspective shifts, all exceeding the corresponding thresholds, and is therefore selected into the structural difference segment set.

[0095] The performance recognition submodule collects students' recognition records of segments in bilingual listening and reading tasks based on the set of structural difference segments. It extracts the correct frequency of sentence pattern recognition, task response time and error expression type code for each segment, compares the recognition frequency with the reference standard data, and analyzes the impact of the combination of response time and error expression changes on the comprehension rhythm to obtain the sentence pattern comprehension lag set.

[0096] First, extract the students' identification records for each segment. Break down the recorded listening and reading behavior logs by timestamp, marking whether students correctly identified the main sentence type or structural pattern after each segment was played. Count the total number of correctly identified sentence patterns in that segment and calculate the correct identification frequency. For example, if segment A contains 9 main sentences and a student correctly identifies 6, the correct identification frequency is 6. Next, read the response time after each listening and reading segment. Record the time spent by the student from finishing listening to starting to answer the question as the task reaction time, timed in seconds. For example, if the student starts answering the question within 6 seconds of finishing listening, it is recorded as a reaction time of 6 seconds. Sort all the segment reaction times in order and compare the trends of time changes in each segment. Then, encode the language errors in the students' answers, setting up an error expression type coding dictionary, such as sentence inversion error as E01, word collocation error as E02, tense usage error as E03, etc. In each section of the answer, the error code and its quantity are marked. For example, if E01 appears once and E03 appears twice in section B, then there are 3 errors, recorded as error type distribution as E01×1 and E03×2. Then, the frequency of correct recognition in each section is compared with the reference standard data. The standard data is set as the average number of sentences correctly recognized by the grade level is 7. If the student only recognizes 5 sentences in the section, it is below the benchmark and is judged as insufficient understanding. Then, the reaction time and error expression type are analyzed together. When the reaction time is greater than the average (e.g., set to 8 seconds) and the total number of error types exceeds 2, it is judged that there is a lag in the understanding of the section. This section number is included in the sentence pattern comprehension lag set. For example, the reaction time of section C is 12 seconds and there are 3 types of errors. It is marked as a lag section. The set of all the section numbers that meet the conditions and their recognition frequency, reaction time and error code is summarized to form the sentence pattern comprehension lag set.

[0097] The structural comparison submodule compares the visual pause distribution, total number of narrative jumps, and comprehension delays for each segment based on sentence structure lag set, using the following formula:

[0098]

[0099] The degree of matching deviation is obtained, and the correspondence with the evolution path of the original structural style of the fragment is analyzed to obtain the structural matching distribution index, where DU s Indicates the degree of matching deviation of structurally different segments under multi-feature comparison, n DU VU represents the total number of structurally different segments. z TU represents the number of visual pauses in the z-th segment, c represents the total number of narrative jumps in the z-th segment, and c represents the difference between the correct frequency of the z-th segment in sentence pattern recognition and the reference standard. z This represents the number of erroneous representation types in the z-th segment.

[0100] The degree of mismatch refers to the difference between the student's actual comprehension performance in a bilingual listening and reading task and the original structural features of the segment (such as visual pauses and narrative jumps) in each structurally different segment. It reflects the student's adaptation to different structural segments: the larger the value, the greater the difference between the student's structural comprehension performance in the segment and the structural features of the segment itself, and the more obvious the deviation from the original structure due to pauses, jumps and incorrect expressions in the comprehension process; the smaller the value, the better the student's comprehension process matches the original structure of the segment, and the smaller the deviation.

[0101] Given that the original data for segment A consists of 3 visual pauses, 2 narrative jumps, 1 fewer correct recognition frequency than the reference standard, and 3 types of incorrect expressions, their normalized values ​​are VU. z =0.60、JU z =0.40 、 RU z =0.20 、 TU z =0.60 ， Substituting the data into the formula, the calculation process for segment A is as follows:

[0102]

[0103] Then, using the original data of segment C, we set it to include 5 visual pauses, 3 narrative jumps, 2 recognition frequency differences, and 2 erroneous expressions. After normalization, this is represented as VU. z =1.00、JU z =0.60 、 RU z =0.40 、 TU z =0.40, then:

[0104]

[0105] Data segment E is set to include 2 visual pauses, 1 narrative jump, 0 recognition frequency difference, and 1 erroneous expression, and is normalized to VU. z =0.40、JU z =0.20 、 RU z =0.00 、 TU z =0.20, calculated as follows:

[0106]

[0107] The original data for segment H consisted of 6 visual pauses, 4 narrative jumps, 3 recognition frequency differences, and 4 erroneous expressions, which were normalized to VU. z =1.20、JU z =0.80 、RU z =0.60 、 TU z =0.80, calculated as follows:

[0108]

[0109] Substitute the matching deviation results of the above four segments into the formula to calculate the average deviation:

[0110]

[0111] This result indicates that when the calculated matching deviation DU s After comparing the value of 0.5109 with the set structural matching baseline interval [0.00, 0.30], it can be determined that the current student's structural understanding performance on the selected structural difference segment has exceeded the upper limit of the basic matching interval and is in the moderate deviation interval [0.30, 0.70]. This indicates that the student has a certain degree of inconsistency and tendency to break in terms of visual recognition, sentence pattern reconstruction and structural maintenance ability. The formula integrates the student's visual and narrative jump load through the square root term, reflects the actual comprehensive degree of structural pressure, and quantifies the difference with the weighted influence of recognition gap and expression error, so as to ensure that the evaluation of the matching degree has a balanced and stable numerical basis.

[0112] Please see Figure 5 The trend tracking module includes:

[0113] The vocabulary recurrence submodule extracts keywords from the detected content based on the structure matching distribution index, counts the frequency of words in each detection task segment, distinguishes the distribution of words in segments with different time sequences, and obtains the vocabulary recurrence density level based on the difference between the number of repetitions and word class coverage within the segment.

[0114] The original text content of each detection segment is read from the task settings, and all content words are extracted using a part-of-speech classification tool, excluding function words such as "the," "a," and "is," forming a preliminary vocabulary list. For example, if the detection segment is "Tom ran to the door and opened it happily," the extracted terms are "Tom," "ran," "door," "opened," and "happily." Other terms are not included in the keyword count. The frequency of each term in the current segment is then counted one by one. For example, if "Tom" appears once and "ran" appears once in the segment, the frequency of each term is recorded as 1. After the count is completed, a repetition check is performed on all segments to determine whether a certain keyword appears repeatedly in other segments of the same detection text. If duplicate records are found, their chronological position is marked, and a distribution correspondence table of terms and segments is constructed. For example, if "door" appears in both segment 1 and segment 4, it is recorded as the 3rd position in segment 1 and the 6th position in segment 4, respectively. The system then determines whether the repeated occurrences span segments with significant time differences, based on the task time sequence set in the detection text. For example, if segment 1 is the beginning scene and segment 4 is the end scene, the word belongs to the time-differentiated distribution term. Subsequently, a joint judgment is made based on the repetition frequency of each keyword and the span of its distribution segments. If a word is repeated more than 3 times and appears in 3 different time logical segments, it is recorded as a high-recurrence term. The frequency of all high-recurrence terms is compared with the total number of low-frequency terms to obtain the degree of difference in the word repetition distribution. Then, combined with the coverage of the word class of the words appearing in each segment, such as whether it is only verbs and nouns, or whether it includes adverbs, adjectives, prepositional phrases, etc., the word class diversity level of each segment is judged. Segments with more than 5 word classes are defined as high-coverage segments, and those with 3 or fewer are low-coverage segments. Finally, a word recurrence density is constructed based on the recurrence frequency and the number of word classes in each segment. The recurrence density level is divided according to the density value, for example, 0.3~0.5 is medium density, and above 0.5 is high density. The word recurrence density level is then output.

[0115] The sentence pattern transformation submodule is based on the vocabulary recurrence density level, counts the number of sentence pattern types in each detected segment, distinguishes the types of changes in sentence pattern structure between adjacent segments, compares the usage of different sentence pattern combinations, summarizes the types of sentence pattern differences and their usage frequency, and obtains the differences in sentence pattern transformation distribution.

[0116] The system detects the use of sentence structures in each paragraph. Sentences from each paragraph are read sequentially, and their types are identified and labeled. Each sentence is classified as a simple sentence, coordinate sentence, passive sentence, conditional sentence, imperative sentence, or exclamatory sentence. The number of each type in the current paragraph is counted. For example, if paragraph 3 contains 4 sentences (2 simple sentences, 1 passive sentence, and 1 conditional sentence), the number of sentence types is recorded as 3. Adjacent paragraphs are then compared horizontally to determine the types of changes in sentence structure usage between paragraphs. For instance, if there are 2 new sentence types between the 3 sentence types in paragraph 3 and the 5 sentence types in paragraph 4, the change type is classified as an addition. If the sentence types used in the two paragraphs are completely identical, it is recorded as a repetition. If all sentence types in paragraph 4 are different from those in paragraph 4, it is classified as an addition. 3. If the sentence pattern change type is recorded, it is considered a completely different type. After recording the sentence pattern change types between all adjacent segments, the total number of times each type of sentence pattern change type appears in the detected text is counted. For example, the "increase type" change appears 4 times, the "decrease type" 2 times, the "repetition type" 3 times, and the "completely different type" 1 time. Then, the number of times each change type is used in high recurrence density segments and low density segments is compared to determine whether a certain change type is concentrated in a certain density interval. If the completely different type of sentence pattern change mainly appears in low recurrence segments, it can be classified as a low density sentence pattern difference type. Otherwise, it is classified as a high density correspondence. Finally, based on the difference type of sentence pattern used, the distribution interval, and the frequency of occurrence, a sentence pattern change distribution table is generated, the proportion of each type of change type in the detection task is counted, and the sentence pattern change distribution difference is output.

[0117] The expression ratio analysis submodule extracts the sentence group length and sentence type number of each segment in the detected content based on the differences in sentence pattern transformation distribution. It then compares each structure according to expression norm parameters, using the following formula:

[0118]

[0119] Calculate the expression density shift amplitude of each segment, integrate the associated expression density data, and obtain the expression trend trajectory group EY. T Where LY represents the sentence group length of the detected segment, DY represents the number of sentence patterns in the detected segment, MY represents the number of standard sentence patterns in the expression norm parameters, and YK represents the structural reference constant under the corresponding detection stage.

[0120] The expression trend trajectory group is a comprehensive distributional parameter that quantifies the proportional relationship between sentence group length (LY), sentence type number (DY), and expression standard (MY) of each segment in the test content, and corrects it with structural reference constant (YK). The final result reflects the changes in students' expression methods, sentence group structure, and degree of conformity with expression standards throughout the entire test task. It reflects the overall change trajectory of students' expression density, sentence type usage, and standard matching status in different segments during the bilingual story retelling test, and provides a basis for analyzing expression coherence, complexity, standardization, and trend changes.

[0121] The sentence group length LY and sentence type number DY of each segment in the detected content are extracted. The sentence group length refers to the number of complete sentences in a single segment. In this embodiment, the sampled segment contains 12 sentences, so LY=12 is set. The sentence type number refers to the number of sentence type types with obvious structural differences within the segment. After labeling, it was confirmed to be 4 types, so DY=4 is set. The number of standard sentence types in the expression norm parameters is set to 5 types according to the language course requirements, so MY=5 is set. The structural reference constant YK is the expected value of the expression structure reference. Based on expert evaluation and the average offset of structurally balanced paragraphs in similar texts, YK=2.2 is set. Because different parameters have different dimensions, they need to be normalized. After normalization, LY=0.75, DY=0.80, MY=1.00, and YK=0.55. Substitute the normalized parameters into the formula:

[0122]

[0123] If the set expression density offset tolerance range is [0, 0.10], then this result indicates that the expression trend trajectory group value EY T =0.05, which is within the set allowable range for expression density deviation. This indicates that the density variation of the current segment's expression structure is relatively close to the standard structure, without drastic fluctuations, and the expression structure remains relatively stable. This value is a basic data point in the expression trend trajectory group, participating in the construction of the structural fluctuation distribution sequence in the multi-segment sequence. The expression deviation values ​​obtained from each segment can be sorted by time series to further deduce the direction, range, and amplitude of expression trend changes. Based on this, they can be integrated into the expression trend trajectory group, thereby supporting the positioning and configuration of subsequent intervention strategies and the adjustment of task rhythm.

[0124] Please see Figure 6 The strategy configuration adjustment module includes:

[0125] The prompt rhythm judgment submodule analyzes the arrangement rhythm of each task in the teaching library based on the expression trend trajectory group, compares the difference with the distribution of graphic intervention prompts, judges the density change of semantic intervention prompts in each task interval, and filters out the abnormal intervals with the best correlation with the turning point of the paragraph to obtain the prompt interval offset.

[0126] Using the trend trajectory group as the analysis benchmark, all task items in the teaching library are extracted, and their setting order and time points are recorded sequentially according to task number to form a teaching task rhythm sequence. For example, the time intervals between tasks 1 to 5 are 3, 4, 2, 5, and 3 minutes, respectively. The variation range of the time interval between adjacent tasks is calculated. Then, the appearance scenarios and positions of the graphic and textual prompts inserted into the tasks are recorded according to their content. The density distribution of graphic and textual intervention prompts is extracted, such as the number of prompts inserted in each paragraph, the position of the prompt in the paragraph, and whether it is a graphic and textual joint prompt. A prompt density sequence is constructed, and the prompt density is paired with the task rhythm sequence one by one according to task number to calculate the difference between the two. The difference is calculated, and if several pre-cues are concentrated in rhythmically dense segments, they are recorded as cues clustered; if they are concentrated in sparsely tasked segments, they are marked as cues dispersed. The time distance between each cue and the task turning point is calculated, and whether it is aligned with the segment turning point is recorded. A distance within 5 seconds is defined as a strong correlation interval, more than 10 seconds as a weak correlation interval, and the middle interval as a normal correlation interval. Areas in all task segments with cues offset distances exceeding 10 seconds are marked as abnormal cues. The task number of the abnormal interval and the corresponding cues offset time are used as the offset interval output. The task segments with the least overlap with the segment turning point and the sudden change in cues density are selected, and the cues interval offset is output.

[0127] The intervention fusion assessment submodule analyzes the distribution of text-image intervention fusion density on the expression nodes of the segment based on the prompt interval offset, determines the correspondence between intervention configuration and node type in each time period of expression, optimizes the intervention distribution order, and obtains the intervention configuration positioning frequency.

[0128] The density of text-image intervention fusion is annotated at the segment level. All segments covered by prompts are read, and the frequency of intervention occurrences in each segment is counted, labeling them as text, image, or combined text-image intervention types. Intervention information and the start and end nodes of segment expression are arranged chronologically to determine if the intervention content falls at key expression nodes, such as the beginning of sentence groups, sentence transitions, or emotional turning points. Logical node types of segment expression are extracted, and node types such as parallel transformations, progressive expressions, and causal explanations are encoded. The frequency of intervention configurations in node types is matched. When the frequency of intervention configurations in a certain type of node is higher than... If an intervention occurs 3 times and other nodes occur 1 time or less, the configuration is marked as centralized. Further, the frequency of all intervention configurations is sorted according to the distribution of node type. The intervention order of high-frequency centralized paragraphs is adjusted, and the text and image interventions are moved before the expression change point or the sentence transformation starting point. The order difference between the original position and the adjusted position is recorded as the intervention migration span. The final position of all adjusted intervention points in the paragraph and the node types they cover are counted to form a positioning matrix of intervention configurations and nodes. The positioning frequency of intervention configurations is output according to the proportion of the number of interventions of each type of node to the total number of intervention configurations.

[0129] The configuration offset calculation submodule analyzes the differences in the density of text-image intervention fusion, sentence density ratio, vocabulary recurrence frequency, and paragraph expression type based on the frequency of intervention configuration positioning, compares the changes in structural arrangement and expression mode, and obtains the magnitude of intervention parameter offset.

[0130] A joint analysis of various structural indicators in the expression process was conducted. First, the corresponding text-image intervention fusion density was extracted for each expression segment, and the ratio of the number of interventions to the length of the expression segment was recorded to determine whether there were abnormal text-image cue densities in each segment. For example, if the segment length was 30 words and the number of interventions was 5, the density was 0.167. If it exceeded 0.2, it was considered a high-density segment, and a cue density threshold of 0.15 was set for differentiation. Then, the sentence structure density ratio was extracted, and the proportion of different sentence structures in each segment was counted. If the segment contained 5 types of sentence structures in 10 sentences, the sentence structure density was 0.5, and a baseline sentence structure density threshold of 0.4 was set. The third item was the word recurrence frequency, which was the ratio of the number of repeated keywords in the segment to the total number of words. If a keyword in a 40-word segment appeared 10 times, the recurrence frequency was 0.25. The fourth item was the word recurrence frequency. The item represents the difference in expression type of the paragraph. It determines whether the current paragraph is narrative, descriptive, explanatory, commentary, or dialogue, and classifies and numbers the expression mode. Each type is labeled with an expression mode code. The change trajectory of expression mode in all paragraphs is calculated. If the expression changes from "explanatory" to "dialogue" and then to "descriptive", it is recorded as two consecutive switches. Then, the combination changes of the above four indicators in each paragraph are compared horizontally to determine whether there are structural mismatch segments with high cue density, low sentence density, high vocabulary repetition but abrupt changes in expression mode. If there is a change trajectory that is inconsistent between structure and expression, the paragraph is marked as a parameter offset segment. The total number of offset segments is counted and its proportion of the total number of segments is calculated. For example, if there are 7 offset segments out of 30 total segments, the intervention parameter offset amplitude is 0.233, and the offset amplitude value is output.

[0131] An adaptive teaching approach for bilingual story comprehension that addresses individual differences includes the following steps:

[0132] S1: Based on role-switching bilingual texts, analyze the number of characters and the order of events. Combined with the content of students' bilingual story retelling, determine the actual distribution of temporal adverbs and character titles used by students in the retelling text. Compare the correspondence between the order of students' retelling and the order of the original text. Based on the data of the level of perception of the story plot, obtain the structural adaptation performance parameters.

[0133] S2: Based on the structural adaptation performance parameters, calculate the number of times students respond to keywords involved in bilingual scenario-based question and answer, analyze the complete performance of syntactic structure in the answer content, compare the semantic preservation in the bilingual translation process, summarize the changes in students' language comprehension characteristics, and obtain the comprehension feature shift.

[0134] S3: Based on understanding the feature offset, select story fragments with significant structural differences in the combined text and image story material library, compare students' performance in sentence pattern recognition and visual comprehension lag in bilingual listening and reading tasks, determine the distribution characteristics of semantic connection breakpoints, and analyze the differences between the story fragment structure and students' comprehension performance to obtain the structure matching distribution index.

[0135] S4: Based on the structure matching distribution index, optimize the expression of segments in the bilingual story retelling test process, count the frequency of word recurrence in the test content, measure the number of sentence pattern transformations, analyze the ratio change of sentence group length and sentence pattern density, summarize the distribution trend of expression mode, and obtain the expression trend trajectory group.

[0136] S5: Based on the expression trend trajectory group, adjust the task arrangement rhythm in the teaching library, determine the setting of semantic intervention prompts in each interval, analyze the matching relationship between the density of graphic intervention fusion and the student's expression trajectory, compare the positioning of intervention configuration in the expression process, and obtain the magnitude of intervention parameter offset.

[0137] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A bilingual story comprehension adaptive teaching system tailored to individual differences, characterized in that: The system includes: The structural adaptation analysis module is based on role-switching bilingual texts. It analyzes the number of roles, the order of events and the content of students' retelling, determines the actual distribution of temporal adverbs and titles of people, compares the order of the retelling with the original text, and combines the perception level classification data to summarize the differences in language structure application and obtain structural adaptation performance parameters. Based on the structural adaptation performance parameters, the understanding feature induction module calculates the number of keyword responses in bilingual plot question-and-answer, analyzes the syntactic structure integrity, compares the semantic preservation during bilingual translation, and, in conjunction with type criteria and transfer strength, summarizes language understanding changes to obtain the understanding feature offset. Based on the aforementioned understanding feature offset, the material matching evaluation module filters out segments with structural differences in the material library, compares the performance of sentence pattern recognition with the lag in visual understanding, determines the distribution of semantic connection interruptions, analyzes the differences in story structure, and obtains the structural matching distribution index. Based on the structure matching distribution index, the expression trend tracking module optimizes the segment expression in bilingual story retelling detection, counts the frequency of word recurrence and the number of sentence pattern transformations, analyzes the changes in sentence group length and density, and obtains the expression trend trajectory group by combining expression normation parameters; Based on the expression trend trajectory group, the strategy configuration adjustment module adjusts the task arrangement rhythm, judges the semantic prompt interval setting, analyzes the matching between text and image intervention and expression trajectory, compares the intervention configuration positioning, summarizes parameter changes, and obtains the intervention parameter offset magnitude. The structural adaptation performance parameters include expression accuracy, language structure complexity, and information integration level; the understanding feature offset includes semantic extraction sensitivity, knowledge transfer amplitude, and understanding stability; the structural matching distribution indicators include material diversity distribution, structural fit, and content association strength; the expression trend trajectory group includes expression pattern changes, content presentation trends, and repetition coherence; and the intervention parameter offset amplitude includes task distribution adjustment amplitude, intervention response sensitivity, and strategy matching level.

2. The bilingual story comprehension adaptive teaching system for individual differences as described in claim 1, characterized in that, The structural adaptation analysis module includes: The role distribution extraction submodule extracts the number of roles and their first appearance order in the text based on the role-converting bilingual text. Combined with the student's retelling text, it identifies the names of the characters one by one. By comparing the appearance order of the character names in the original text and the retelling text, it counts the frequency of the misalignment of the first appearance order of the characters between the two texts and obtains the difference in the character order. The temporal distribution comparison submodule calls the role order difference quantity to detect all temporal adverbs in the student's retelling text, distinguishes their distribution range and frequency of use by paragraph, arranges the adverbs used for each event sequence in the original text to correspond, and compares the paragraph coverage and mutual exclusion distribution of time markers in the two texts to obtain the event sequence misalignment rate. The structural adaptation induction submodule analyzes the frequency of conjunctions, syntactic nesting levels, subject substitution frequency, and sentence group transition density in the students' retelling content based on the event sequence misalignment rate. It also combines the plot perception level classification data and performs type interval matching on the features to obtain structural adaptation performance parameters.

3. The bilingual story comprehension adaptive teaching system for individual differences as described in claim 1, characterized in that, The feature induction module includes: Based on the aforementioned structural adaptation performance parameters, the keyword response submodule statistically analyzes the responses of students to all plot keywords in the bilingual plot question-and-answer task, distinguishes the different task stages, categorizes the position and repetition frequency of keywords in each stage, and accumulates and sums the number of keyword responses to obtain the total number of keyword responses. The syntactic structure detection submodule decomposes the sentences of the response content based on the total number of keyword responses, retrieves the completeness of the subject, predicate, object and modifiers, and classifies the syntactic structure coverage performance according to the sentence construction type standard to obtain the syntactic structure integrity level coefficient. The semantic transfer comparison submodule extracts bilingual translation content based on the syntactic structure integrity level coefficient, decomposes the target language vocabulary and semantic information, compares the source language information retention performance, and obtains the understanding feature offset by combining the transfer strength and sentence change index.

4. The bilingual story comprehension adaptive teaching system for individual differences as described in claim 1, characterized in that, The material matching and evaluation module includes: Based on the understanding feature offset, the segment filtering submodule obtains the segment structure features of the combined text and image story material library, collects the number of sentence patterns, the total number of information levels and the frequency of character perspective changes for each segment, analyzes the structural span based on the differences in sentence patterns, information levels and perspective frequency, and obtains a set of structural difference segments by filtering segments that meet the conditions. The performance recognition submodule collects students' recognition records of segments in bilingual listening and reading tasks based on the set of structural difference segments, extracts the correct frequency of sentence pattern recognition, task response time and error expression type code for each segment, compares the recognition frequency with the reference standard data, analyzes the impact of the combination of response time and error expression on the rhythm of understanding, and obtains the sentence pattern comprehension lag set. The structural comparison submodule, based on the lag set of the sentence pattern comprehension, compares the visual pause distribution, total number of narrative jumps, and comprehension delay performance of each segment to obtain the degree of matching deviation, analyzes the degree of correspondence with the evolution path of the original structural style of the segment, and obtains the structural matching distribution index.

5. The bilingual story comprehension adaptive teaching system for individual differences as described in claim 1, characterized in that, The expression trend tracking module includes: The vocabulary recurrence submodule extracts keywords from the detected content based on the structure matching distribution index, counts the frequency of occurrence of words in each detection task segment, distinguishes the distribution of words in segments with different time sequences, and obtains the vocabulary recurrence density level based on the difference between the number of repetitions and word class coverage within the segment. Based on the vocabulary recurrence density level, the sentence pattern transformation submodule counts the number of sentence pattern types in each detected segment, distinguishes the types of changes in sentence pattern structure between adjacent segments, compares the usage of different sentence pattern combinations, summarizes the types of sentence pattern differences and their usage frequency, and obtains the differences in sentence pattern transformation distribution. The expression ratio analysis submodule extracts the sentence group length and sentence type number of each segment in the detected content based on the differences in sentence pattern transformation distribution. It compares each structure according to the expression standard parameters, calculates the expression density offset of each segment, integrates the related expression density data, and obtains the expression trend trajectory group.

6. The bilingual story comprehension adaptive teaching system for individual differences as described in claim 1, characterized in that, The strategy configuration adjustment module includes: The prompt rhythm judgment submodule analyzes the arrangement rhythm of each task in the teaching library based on the expression trend trajectory group, compares it with the difference in the distribution of graphic intervention prompts, judges the density change of semantic intervention prompts in each task interval, and filters out the abnormal intervals with the best correlation with the turning point of the paragraph to obtain the prompt interval offset. Based on the prompt interval offset, the intervention fusion evaluation submodule analyzes the distribution of the text-image intervention fusion density on the segment expression nodes, determines the correspondence between the intervention configuration and the node type in each time period of the expression, optimizes the intervention distribution order, and obtains the intervention configuration positioning frequency. The configuration offset calculation submodule analyzes the differences in the density of text-image intervention fusion, sentence density ratio, vocabulary recurrence frequency, and paragraph expression type based on the frequency of the intervention configuration positioning, compares the changes in structural arrangement and expression mode, and obtains the magnitude of the intervention parameter offset.

7. An adaptive bilingual story comprehension teaching method tailored to individual differences, characterized in that: The method is used to implement the bilingual story comprehension adaptive teaching system oriented towards individual differences as described in any one of claims 1-6, and includes the following steps: S1: Based on role-switching bilingual texts, analyze the number of characters and the order of events. Combined with the content of students' bilingual story retelling, determine the actual distribution of temporal adverbs and character titles used by students in the retelling text. Compare the correspondence between the order of students' retelling and the order of the original text. Based on the data of the level of perception of the story plot, obtain the structural adaptation performance parameters. S2: Based on the structural adaptation performance parameters, calculate the number of times students respond to keywords involved in the bilingual scenario question and answer process, analyze the complete performance of syntactic structure in the answer content, compare the semantic preservation in the bilingual translation process, summarize the changes in students' language comprehension characteristics, and obtain the comprehension feature offset. S3: Based on the aforementioned understanding feature offset, screen story fragments with significant structural differences within the combined text and image story material library, compare students' performance in sentence pattern recognition and visual comprehension lag in bilingual listening and reading tasks, determine the distribution characteristics of semantic connection breakpoints, and analyze the differences between the story fragment structure and students' comprehension performance to obtain structural matching distribution indicators. S4: Based on the structure matching distribution index, optimize the expression of segments in the bilingual story retelling test process, count the frequency of word recurrence in the test content, measure the number of sentence pattern transformations, analyze the ratio change of sentence group length and sentence pattern density, summarize the distribution trend of expression mode, and obtain the expression trend trajectory group. S5: Based on the expression trend trajectory group, adjust the task arrangement rhythm in the teaching library, determine the setting of semantic intervention prompts in each interval, analyze the matching relationship between the density of graphic intervention fusion and the student's expression trajectory, compare the positioning of intervention configuration in the expression process, and obtain the magnitude of intervention parameter offset.

Images