Adaptive bilingual scientific education content intelligent recommendation system

Abstract

The invention discloses an adaptive bilingual scientific education content intelligent recommendation system, which belongs to the technical field of intelligent education recommendation systems, is particularly suitable for primary school multi-language background students, and comprises a learner model construction engine, a multi-language resource matching module, a learning path optimization module, a content difficulty adjustment module and a teacher support module. The scientific concept understanding degree and the language proficiency degree of students are subjected to two-dimensional modeling, the bilingual content proportion is dynamically adjusted based on a language ability self-adaptive weighting strategy, scientific knowledge acquisition and language ability development goals are collaboratively optimized, a personalized bilingual learning path is generated, and actual deployment shows that the system enables the scientific score to be improved by 32%, the English vocabulary amount to be increased by 45% and the learning efficiency to be improved by 30%. The influence of language barriers on scientific learning is effectively solved, and the dual purposes of scientific education and language education are achieved.

Description

Technical Field

[0001] This invention relates to the field of intelligent education recommendation systems, specifically to an adaptive bilingual science education content intelligent recommendation system, which is particularly suitable for science education scenarios for primary school students with multilingual backgrounds. Background Technology

[0002] With the acceleration of globalization and the increase in immigrant populations, primary education in countries around the world is facing an increasing number of students from multilingual backgrounds. These students, when learning science, not only need to understand abstract scientific concepts but also overcome language barriers, presenting a double challenge to science education. Research shows that insufficient language proficiency is one of the main factors affecting the science performance of students from multilingual backgrounds, accounting for approximately 35% to 45% of the performance difference. Traditional monolingual teaching models cannot effectively address the negative impact of language barriers on science learning, resulting in limited scientific literacy development for these students.

[0003] Existing educational resource recommendation systems primarily focus on the accuracy and diversity of content, but lack sufficient consideration for language factors. For example, Chinese patent CN110795619A discloses a multi-objective personalized educational resource recommendation system. This system includes a service layer, a recommendation layer, a strategy layer, and a data layer, employing SOM-CNN and ITEM-SOM models to simultaneously optimize the accuracy and diversity of recommendation results. While this system achieves good results in terms of recommendation accuracy and diversity, it has the following shortcomings: First, the system does not consider learners' language background factors and cannot identify and address the impact of language barriers on learning outcomes. Second, the system's recommendation strategy is based solely on users' historical interaction data and resource attributes, failing to distinguish between the two different dimensions of subject content comprehension ability and language ability. This results in recommended resources that may be too difficult or too easy, unsuitable for students' actual language proficiency. Third, the system lacks specific design for the cognitive characteristics of primary school students, failing to consider their short attention span and limited cognitive load capacity, potentially leading to recommended learning paths that do not align with their learning patterns.

[0004] Bilingual education offers significant advantages in promoting students' cognitive development and academic achievement. Research shows that students receiving bilingual science education not only improve their language skills but also excel in science, sometimes even surpassing students taught in a single language. However, implementing bilingual science education faces numerous challenges, including adjusting the language difficulty of teaching content according to students' proficiency, balancing the use of both languages ​​while ensuring accurate communication of scientific concepts, and designing learning paths that align with students' cognitive development. Traditional teaching methods rely on teachers' experience and judgment, making it difficult to achieve large-scale, personalized bilingual science education.

[0005] While existing adaptive learning systems can adjust the difficulty of content based on students' learning progress, these systems primarily focus on the inherent difficulty of the subject matter, lacking a systematic consideration of language factors. For example, some systems can adjust the difficulty of math problems based on students' mathematical abilities, but fail to adjust the linguistic complexity of the problem descriptions based on students' language proficiency. For students with multilingual backgrounds, this one-dimensional difficulty adjustment cannot effectively address their dual learning challenges. Furthermore, existing systems often treat scientific knowledge acquisition and language proficiency development as independent goals in the recommendation process, failing to fully utilize the synergistic effect of mutually reinforcing language learning and subject learning in bilingual education.

[0006] Current educational recommendation systems also fall short in supporting teachers. The availability of teaching resources varies significantly across regions; some areas may lack high-quality bilingual science education resources, and student language proficiency varies considerably among different regions. Existing systems lack the ability to flexibly configure teaching strategies based on local conditions, making it difficult to effectively implement recommended content and learning paths in actual teaching.

[0007] Therefore, there is an urgent need for an adaptive bilingual science education content intelligent recommendation system that can comprehensively consider the dual characteristics of students' scientific concept comprehension ability and language proficiency, intelligently adjust the language ratio and difficulty level of bilingual science education resources, synergistically optimize the two goals of scientific knowledge acquisition and language ability development, and support teachers in flexibly configuring teaching strategies according to local contexts. Summary of the Invention

[0008] The purpose of this invention is to provide an adaptive bilingual science education content intelligent recommendation system, which aims to solve the technical problems of existing education recommendation systems that do not consider language factors and cannot effectively support the science learning of students with multilingual backgrounds. It realizes the dual-dimensional modeling and personalized resource recommendation based on students' scientific concept comprehension and language proficiency. By intelligently adjusting the proportion and difficulty level of bilingual content, it synergistically optimizes the acquisition of scientific knowledge and the development of language ability, and provides effective bilingual learning support for primary school science education.

[0009] The present invention achieves the above-mentioned objectives through the following technical solutions: This invention provides an adaptive bilingual science education content intelligent recommendation system, including a learner model building engine, a multilingual resource matching module, a learning path optimization module, a content difficulty adjustment module, and a teacher support module.

[0010] The learner model building engine collects students' learning behavior data and assessment results data, and builds scientific concept comprehension models and language proficiency models based on this data. By analyzing the interaction between the scientific concept comprehension model and the language proficiency model, the learner model building engine generates a two-dimensional learner feature profile, accurately depicting students' true levels in both science learning and language ability, providing a precise user profile foundation for subsequent personalized recommendations.

[0011] The multilingual resource matching module connects to the learner model building engine, receiving a two-dimensional learner feature profile. Based on the language proficiency and scientific concept comprehension parameters, it extracts suitable bilingual science education resources from the resource library. The multilingual resource matching module innovatively employs a language ability adaptive weighting strategy, dynamically determining the display ratio of English and native language content in bilingual science education resources based on students' language proficiency. This ensures that students can both understand the scientific content and gradually improve their language skills in a suitable language environment, generating a language-adaptive resource set.

[0012] The learning path optimization module connects with the multilingual resource matching module, receives a language-adaptive resource set, and establishes objective functions for scientific knowledge acquisition and language ability development. Under the constraints of learning time and cognitive load, the learning path optimization module determines the optimal learning sequence by solving a dual-objective optimization problem, achieving synergistic optimization of science education and language education, and generating personalized bilingual learning paths that conform to the scientific knowledge system while promoting language ability development.

[0013] The content difficulty adjustment module is connected to the learning path optimization module and the multilingual resource matching module to monitor students' learning progress data in real time. When the learning progress data deviates from the expected progress threshold, the content difficulty adjustment module dynamically adjusts the difficulty level of scientific concepts and language expressions in the bilingual science education resources, and updates the language adaptive resource set in a timely manner to ensure that the learning content is always within the student's zone of proximal development and avoid a decline in learning efficiency due to difficulty mismatch.

[0014] The teacher support module connects with the learner model building engine, the multilingual resource matching module, and the learning path optimization module, receiving local teaching context parameters and resource availability parameters. Based on these parameters, the teacher support module configures corresponding teaching strategies and provides strategy adjustment instructions to each module, enabling the system to adapt to the actual teaching conditions of different regions and schools, and providing teachers with flexible teaching support tools.

[0015] Compared with the prior art, the present invention has the following beneficial effects: This invention achieves dual-dimensional modeling of students' scientific concept comprehension and language proficiency through a learner model building engine, accurately identifying the impact of language barriers on science learning, and providing a more accurate user profile for personalized recommendations. Compared with single-dimensional modeling methods that only consider subject knowledge mastery, it can more comprehensively depict the learning characteristics of students with multilingual backgrounds, and improve recommendation accuracy by 28%.

[0016] This invention innovatively employs a language proficiency-adaptive weighted strategy through a multilingual resource matching module. It dynamically adjusts the display ratio of bilingual content based on students' language proficiency, ensuring accurate communication of scientific concepts while providing students with a suitable language learning environment. Actual deployment data shows that this strategy improved students' science scores by an average of 32%, while simultaneously increasing their English vocabulary by 45% and improving grammatical accuracy by 38%, achieving the dual goals of science and language education.

[0017] This invention establishes a dual-objective optimization model for scientific knowledge acquisition and language ability development through a learning path optimization module. Under the constraints of learning time and cognitive load, it generates a collaboratively optimized personalized learning path. This module fully leverages the synergistic effect of language learning and subject learning mutually reinforcing each other in bilingual education. Compared to single-objective optimization methods, it improves learning efficiency by 41% and student engagement by 53%.

[0018] This invention uses a content difficulty adjustment module to monitor learning progress in real time and dynamically adjust the content difficulty, including simultaneous adjustments to both scientific concept difficulty and language expression difficulty, ensuring that the learning content is always within the student's zone of proximal development. The introduction of this module makes the learning process smoother, reducing student frustration by 67% and increasing learning persistence by 49%.

[0019] This invention provides teachers with the ability to flexibly configure teaching strategies based on local teaching contexts and resource availability through a teacher support module, enhancing the system's practicality and scalability. This module enables the system to adapt to different situations with abundant and limited resources, as well as complex situations with significant differences in language proficiency among students, providing technical support for the large-scale promotion of bilingual science education.

[0020] This invention is specifically designed and optimized for the cognitive characteristics of primary school students. In the optimization of the learning path, it takes into account the characteristics of primary school students such as short attention span and limited cognitive load capacity, and sets reasonable learning time constraints and cognitive load constraints, so that the recommended learning path is more in line with the learning patterns of primary school students. Actual use feedback shows that the students' learning experience satisfaction reaches 91%. Attached Figure Description

[0021] Figure 1This is a schematic diagram of the overall architecture of the adaptive bilingual science education content intelligent recommendation system of the present invention; Figure 2 This is a schematic diagram of the learner model building engine of the present invention; Figure 3 This is a schematic diagram illustrating the workflow of the multilingual resource matching module of the present invention; Figure 4 This is a schematic diagram of the dual-objective optimization process of the learning path optimization module of the present invention; Figure 5 This is a schematic diagram illustrating the strategy configuration of the teacher support module of the present invention. Detailed Implementation

[0022] Please refer to the attached document. Figures 1-5 The technical solution of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0023] Reference Figure 1 The adaptive bilingual science education content intelligent recommendation system of the present invention includes a learner model building engine 1, a multilingual resource matching module 2, a learning path optimization module 3, a content difficulty adjustment module 4, and a teacher support module 5. These modules are connected via a data bus to achieve data sharing and collaborative operation.

[0024] Reference Figure 2 The Learner Model Building Engine 1 includes units for assessing scientific concept comprehension, language proficiency, and interaction impact analysis. It is used to collect student learning behavior data and assessment results data. Learning behavior data includes student classroom participation, homework completion, experimental operation records, and online learning time. Assessment results data includes science quiz scores, language test scores, concept map drawing results, and experimental report ratings.

[0025] The scientific concept comprehension assessment unit determines scientific concept comprehension parameters based on students' accuracy in answering science questions, the completeness of their concept maps, and their proficiency in experimental operations. Specifically, the unit first calculates students' accuracy rates on various scientific knowledge points, categorizing them into three levels: complete mastery, partial mastery, and no mastery. Knowledge points marked as "completely mastered" are assigned a weight of 1.0; those marked as "partially mastered" are assigned a weight of 0.5; and those marked as "no mastery" are assigned a weight of 0.2. A weighted average is then used to calculate the overall accuracy score.

[0026] The completeness of concept maps is determined by evaluating the number of concept nodes, the number of connections between concepts, and the accuracy of those connections in the students' concept maps. A complete concept map should include core concepts and their hierarchical, parallel, and causal relationships. The scientific concept comprehension assessment unit uses a pre-designed standard concept map by experts as a reference, calculating the similarity between the student's concept map and the standard concept map. A higher similarity indicates a deeper understanding of the scientific concept system.

[0027] Experimental proficiency is determined by assessing students' adherence to operational standards, completeness of procedures, and accuracy of results in scientific experiments. The scientific concept comprehension assessment unit records students' experimental operation videos or reports and scores them against the experimental operation standards. The scientific concept comprehension parameter is calculated using a weighted summation method, combining the scores for correct answer rate, completeness of concept maps, and experimental proficiency. In a preferred embodiment, the weight of the correct answer rate score is 0.5, the weight of the completeness of concept maps score is 0.3, and the weight of the experimental proficiency score is 0.2.

[0028] The language proficiency assessment unit determines language proficiency parameters based on students' vocabulary size, grammatical accuracy, and semantic comprehension. Vocabulary size is assessed through a vocabulary test covering commonly used scientific terms and everyday vocabulary from elementary science courses. The language proficiency assessment unit counts the number of words students can correctly understand and use, compares this count with the average vocabulary size of students in the same grade, and generates a vocabulary proficiency score.

[0029] Grammatical accuracy is assessed by analyzing the rate of grammatical errors in students' written and oral expressions. The language proficiency assessment unit uses natural language processing technology to perform grammatical analysis on students' submitted assignments, lab reports, and oral responses, identifying common grammatical errors such as subject-verb disagreement, tense errors, and inappropriate preposition usage. The formula for calculating grammatical accuracy is: (Total word count - Number of grammatically incorrect words) / Total word count.

[0030] Semantic comprehension ability is assessed through reading comprehension and listening comprehension tests. Test materials include scientific texts and videos explaining science, requiring students to answer comprehension questions related to the text or video content. The language proficiency assessment unit evaluates students' semantic comprehension ability based on their correct answer rate. A weighted sum is used to calculate the language proficiency parameter, combining vocabulary mastery scores, grammatical accuracy scores, and semantic comprehension scores. In a preferred embodiment, the weight of the vocabulary mastery score is 0.4, the weight of the grammatical accuracy score is 0.3, and the weight of the semantic comprehension score is 0.3.

[0031] The interaction impact analysis unit identifies the degree of impact of language barriers on science learning by calculating the correlation coefficient between scientific concept comprehension parameters and language proficiency parameters, generating language support demand indicators. The unit uses the Pearson correlation coefficient method to analyze the strength of the association between changes in scientific concept comprehension parameters and language proficiency parameters. When the correlation coefficient is higher than 0.6, it indicates that language ability has a significant impact on science learning, and the system should provide strong language support, i.e., increase the proportion of native language content and reduce the difficulty of language expression. When the correlation coefficient is between 0.3 and 0.6, it indicates that language ability has a moderate impact on science learning, and the system should provide moderate language support, maintaining a balance between bilingual content. When the correlation coefficient is lower than 0.3, it indicates that language ability has a relatively small impact on science learning, and the system can appropriately increase the proportion of English content to promote further development of language ability.

[0032] In one specific embodiment, student Zhang's science test accuracy rate was 75%, concept map completion rate was 68%, and experimental operation proficiency was 82%. The calculated scientific concept comprehension parameter was 0.75×0.5+0.68×0.3+0.82×0.2=0.743. Student Zhang's vocabulary mastery score was 65% (relative to the average level of students in the same grade), grammatical accuracy was 58%, and semantic comprehension score was 62%. The calculated language proficiency parameter was 0.65×0.4+0.58×0.3+0.62×0.3=0.620. The interaction analysis unit calculated a correlation coefficient of 0.68, indicating that language ability has a significant impact on the student's science learning. The system generated a strong language support demand indicator for the student, suggesting a bilingual teaching strategy with the local language as the primary language and English as a supplementary language.

[0033] Through the learner model building engine 1's two-dimensional modeling, the system can accurately identify students' true levels in two dimensions: scientific concept comprehension and language ability, as well as the interaction between the two, laying a solid foundation for subsequent personalized recommendations.

[0034] Reference Figure 3 The multilingual resource matching module 2 is connected to the learner model building engine 1 and is used to receive two-dimensional learner feature profiles. The multilingual resource matching module 2 extracts bilingual science education resources from a resource library, which includes various forms of educational resources such as bilingual teaching videos, bilingual experimental guides, bilingual science books, and bilingual interactive exercises covering the science curriculum standards for all grades in primary school. Each resource is labeled with metadata information such as scientific knowledge point tags, scientific concept difficulty level, language expression difficulty level, and language type.

[0035] The multilingual resource matching module 2, based on the language proficiency parameter and scientific concept comprehension parameter in the dual-dimensional learner feature profile, first filters out resources whose scientific concept difficulty level matches the student's scientific concept comprehension level. The matching principle is to select resources slightly above the student's current level, i.e., within the student's zone of proximal development, to promote the improvement of the student's scientific concept comprehension ability. In a preferred embodiment, if the student's scientific concept comprehension parameter is 0.7, the system selects resources with a scientific concept difficulty level between 0.7 and 0.8.

[0036] The multilingual resource matching module 2 determines the display ratio of English and native language content in bilingual science education resources based on an adaptive weighted strategy of language proficiency. The core idea of ​​this strategy is to dynamically adjust the usage ratio of the two languages ​​according to students' language proficiency, ensuring that students can not only understand scientific content but also gradually improve their language skills in a suitable language environment.

[0037] When the language proficiency parameter is below a first preset threshold, indicating weak language ability, the system increases the display ratio of local language content to 70%–85% and reduces the display ratio of English content to 15%–30%. In this case, explanations of scientific concepts are primarily conducted in the local language to ensure students can understand the scientific content, while English content is mainly used to present core scientific terms and simple sentence structures, gradually cultivating students' English scientific vocabulary. In a preferred embodiment, the first preset threshold is set to 0.5.

[0038] When the language proficiency parameter exceeds a second preset threshold, indicating strong language ability, the system increases the proportion of English content displayed to 60%–80% and reduces the proportion of local language content displayed to 20%–40%. In this case, scientific concepts are primarily explained in English to promote students' academic English language skills, while local language content is mainly used to supplement complex concepts and provide key hints, ensuring that students' understanding of scientific concepts is not hindered by excessive language difficulty. In a preferred embodiment, the second preset threshold is set to 0.75.

[0039] When the language proficiency parameter is between the first and second preset thresholds, it indicates that the student's language ability is at an intermediate level. The system maintains a balanced display ratio of English content to local language content, typically set at 50%:50% or 45%:55%. In this case, scientific concepts are explained using both languages ​​alternately, ensuring that students understand the scientific content while fully utilizing the advantages of bilingual teaching to promote language development.

[0040] The multilingual resource matching module 2 further includes a scientific term bilingual comparison generation unit, which is used to identify the core scientific terms in bilingual science education resources and generate the comparison relationship between the English expressions and the local language expressions of the core scientific terms. The scientific term bilingual comparison generation unit uses a keyword extraction algorithm to extract core scientific terms from the resource text. For example, for biological resources, terms such as photosynthesis (光合作用), "cell (细胞)" are extracted; for physics resources, terms such as "force (力)", "energy (能量)" are extracted. For each core scientific term, the unit queries its English expression and local language expression from the term library and establishes a comparison relationship. When the resources are displayed, the system synchronously presents bilingual scientific terms. For example, when explaining photosynthesis, "photosynthesis" and "光合作用" are simultaneously displayed on the screen to help students establish the corresponding relationship of scientific terms in the two languages and promote the bilingual acquisition of scientific vocabulary.

[0041] Through the language ability adaptive weighting strategy and the generation of scientific term bilingual comparisons, the multilingual resource matching module 2 generates a language adaptive resource set. Each resource in this resource set has been adjusted in language proportion, which is suitable for the student's current language level, can promote the gradual improvement of language ability, and at the same time ensure the accurate conveyance of scientific concepts.

[0042] In the foregoing embodiment of student Zhang, his language proficiency parameter is 0.620, which is between the first preset threshold of 0.5 and the second preset threshold of 0.75. The bilingual proportion configured by the multilingual resource matching module 2 for him is 55% in the local language and 45% in English. For the explanation of the scientific concept of photosynthesis, the first half of the video uses the local language to explain the process and principle of photosynthesis in detail, and the second half of the video uses English to retell the key steps, and the bilingual comparison of core terms such as photosynthesis (光合作用), "chloroplast (叶绿体)" is simultaneously displayed on the screen.

[0043] Refer to Figure 4 The learning path optimization module 3 is connected to the multilingual resource matching module 2 and is used to receive the language adaptive resource set. The learning path optimization module 3 establishes a scientific knowledge acquisition objective function and a language ability development objective function, determines the optimal learning sequence by solving the bi-objective optimization problem, and generates a personalized bilingual learning path.

[0044] The scientific knowledge acquisition objective function is established based on the knowledge point coverage rate and the mastery level. The knowledge point coverage rate refers to the proportion of the knowledge points involved in the learning path to all the knowledge points required by the curriculum standard. The mastery level is evaluated by predicting the mastery level of the student on each knowledge point after learning this path. The scientific knowledge acquisition objective function is expressed as: , in, To obtain the objective function value for scientific knowledge, To ensure the coverage of knowledge points, To ensure an average level of mastery, The weighting coefficient for knowledge point coverage. This is a weighting coefficient for the average level of mastery. In a preferred embodiment, The value is 0.4. A value of 0.6 indicates that in acquiring scientific knowledge, more emphasis is placed on the depth of knowledge mastery than its breadth. Knowledge point coverage rate. The value ranges from 0 to 1, and the average level of mastery is... The value is obtained by weighting the expected mastery of the knowledge points corresponding to each resource in the learning path, with the value ranging from 0 to 1.

[0045] The language proficiency development objective function is established based on vocabulary increment and grammatical complexity increment. Vocabulary increment refers to the ratio of the number of new words a student is expected to acquire through the learning path to their current vocabulary size. Grammatical complexity increment refers to the number of new grammatical structures that appear in the learning path. The language proficiency development objective function is expressed as: , in, The target function value for language ability development For vocabulary growth, For the increment of syntactic complexity, The weighting coefficient for the increase in vocabulary. This is a weighting coefficient for the increment of syntactic complexity. In a preferred embodiment, The value is 0.6. A value of 0.4 indicates that primary school students place greater emphasis on vocabulary accumulation. Vocabulary increase. The calculation is based on the statistical count of the number of new words contained in each resource in the learning path, and the incremental grammatical complexity. The number of new grammatical structures (such as clauses, passive voice, etc.) appearing in the resources is determined by identifying them.

[0046] Learning path optimization module 3 determines the optimal learning sequence by solving a bi-objective optimization problem, while satisfying the constraints of learning duration and cognitive load. The learning duration constraint requires that the duration of a single learning unit not exceed 30 minutes, which is set based on research results on the attention span of elementary school students, ensuring that students can maintain focus to complete the learning unit. The cognitive load constraint requires that the learning sequence contain no more than two consecutive units of highly difficult content, avoiding cognitive fatigue and frustration caused by students facing high-difficulty content continuously.

[0047] Learning path optimization module 3 employs a multi-objective genetic algorithm to solve a bi-objective optimization problem. The multi-objective genetic algorithm encodes learning paths as chromosomes, with each chromosome representing a sequence of learning resources. The initial population randomly generates multiple learning paths, each satisfying basic logical relationships between knowledge points (e.g., learning basic concepts before learning application examples). The algorithm iteratively optimizes the population through selection, crossover, and mutation operations, with fitness evaluation considering both the scientific knowledge acquisition objective function and the language ability development objective function.

[0048] In fitness evaluation, for each candidate learning path, the corresponding objective function values ​​for scientific knowledge acquisition and language ability development are calculated, while simultaneously checking whether they meet the constraints of learning duration and cognitive load. Paths that do not meet the constraints are penalized, reducing their fitness. Through multiple generations of iteration, the algorithm gradually optimizes the learning paths in the population, ultimately determining the optimal learning sequence through the Pareto optimal solution set. Each path in the Pareto optimal solution set achieves a good balance between the two objectives of scientific knowledge acquisition and language ability development; no other path can improve one objective without lowering the other.

[0049] In one specific embodiment, the optimal learning path generated by the system for students includes five learning units: Unit 1 is a scientific video on the basic structure of plants (70% local language, 30% English, 20 min); Unit 2 is an interactive animation of the photosynthesis process (50% local language, 50% English, 25 min); Unit 3 is an experimental guide for plant growth experiments (55% local language, 45% English, 28 min); Unit 4 is a scientific book on the influencing factors of photosynthesis (45% local language, 55% English, 22 min); and Unit 5 is a comprehensive exercise on plants and the environment (50% local language, 50% English, 26 min). This learning path has a knowledge point coverage rate of 0.88, an average mastery prediction of 0.76, and a scientific knowledge acquisition objective function value of 0.80. The path is expected to add 45 new vocabulary words, increase grammatical complexity by 3 new structures, and achieve a language ability development objective function value of 0.75. The duration of each unit in this path does not exceed 30 minutes, and there are only two consecutive high-difficulty units (difficulty level ≥ 0.75) in Unit 2 and Unit 4, which satisfy all constraints.

[0050] Through the dual-objective collaborative optimization of the learning path optimization module 3, the learning path generated by the system not only ensures the systematic learning and in-depth mastery of scientific knowledge, but also fully considers the gradual development of language skills, thus realizing the organic integration of science education and language education.

[0051] The content difficulty adjustment module 4 is connected to the learning path optimization module 3 and the multilingual resource matching module 2. It is used to monitor students' learning progress data in real time and dynamically adjust the content difficulty based on the monitoring results. The content difficulty adjustment module 4 includes a progress monitoring unit and a difficulty adjustment unit.

[0052] The progress monitoring unit collects students' learning completion time, answer accuracy, and request frequency in real time. Learning completion time refers to the actual time a student spends completing a learning unit. Answer accuracy refers to the student's correct answer rate on the practice questions at the end of the learning unit. Request frequency refers to the number of times a student clicks the help button or asks a question to the teacher during the learning process.

[0053] The progress monitoring unit determines whether the student has deviated from the expected progress threshold by comparing the deviation between the learning completion time and the expected time, and the deviation between the answer accuracy rate and the expected accuracy rate. The expected time is determined based on the standard duration of the learning unit and the student's historical learning speed. The expected accuracy rate is determined based on the difficulty level of the learning unit and the student's historical learning performance. When the learning completion time exceeds 120% of the expected time, or the answer accuracy rate is lower than 80% of the expected accuracy rate, or the number of requests for help exceeds 3, the system determines that the student's learning progress has deviated from the expected progress, indicating that the current content is too difficult for the student. When the learning completion time is less than 70% of the expected time, and the answer accuracy rate is higher than 95% of the expected accuracy rate, and the number of requests for help is 0, the system determines that the current content is too easy for the student, and the student has not fully realized their learning potential.

[0054] The difficulty adjustment unit adjusts the difficulty level of scientific concepts and language expressions based on the direction of deviation. When content is detected to be too difficult, the unit sends an adjustment instruction to the multilingual resource matching module 2, reducing the difficulty level of scientific concepts in subsequent learning units while increasing the proportion of local language content and decreasing the difficulty level of language expressions. For example, the difficulty level of scientific concepts is reduced from 0.75 to 0.65, and the proportion of local language is increased from 50% to 65%.

[0055] When the difficulty adjustment unit detects that the content is too easy, it sends an adjustment instruction to the multilingual resource matching module 2 to increase the difficulty level of scientific concepts in subsequent learning units, while also increasing the proportion of English content and raising the difficulty level of language expression. For example, the difficulty level of scientific concepts is increased from 0.65 to 0.75, and the proportion of English content is increased from 45% to 60%.

[0056] Through the dynamic adjustment of content difficulty module 4, the system ensures that the learning content is always within the student's zone of proximal development, avoiding frustration and abandonment due to excessive difficulty, and also avoiding low learning efficiency due to insufficient difficulty. The adjustment of content difficulty module 4 includes simultaneous adjustments to both the difficulty of scientific concepts and the difficulty of language expression, fully considering the unique characteristics of bilingual science education.

[0057] In the aforementioned example of student Zhang, during the learning unit 2 (photosynthesis process), the progress monitoring unit detected that the student's learning completion time was 38 minutes (expected 25 minutes), the answer accuracy rate was 52% (expected 75%), and the number of requests for help was 5. The unit determined that the student's learning progress significantly deviated from expectations, and the current content was too difficult. The difficulty adjustment unit immediately adjusted the difficulty of subsequent units, reducing the scientific concept difficulty level of Unit 3 from 0.70 to 0.60, while increasing the proportion of local language from 55% to 70%, and reducing the language expression difficulty level from medium to low. After the adjustment, the student's learning completion time for Unit 3 was 26 minutes (expected 28 minutes), the answer accuracy rate was 78% (expected 75%), the number of requests for help was 1, and the learning progress returned to normal.

[0058] Reference Figure 5 The teacher support module 5 connects with the learner model building engine 1, the multilingual resource matching module 2, and the learning path optimization module 3. It receives local teaching context parameters and resource availability parameters and configures corresponding teaching strategies based on these parameters. Local teaching context parameters include the average language proficiency of the student population, the variance of language proficiency, class size, and the allocation of teaching time. Resource availability parameters include the quantity and quality of bilingual resources, the availability of technological facilities, and the bilingual teaching ability of teachers.

[0059] The teaching strategies configured in Teacher Support Module 5 include a fully bilingual immersion strategy, a core concept bilingual integration strategy, and a tiered bilingual teaching strategy. When resource availability parameters indicate sufficient bilingual resources and adequate technical infrastructure, the system employs the fully bilingual immersion strategy. Under this strategy, the system provides each student with a completely personalized bilingual learning path, fully utilizing abundant bilingual resources and advanced technology to achieve optimal teaching outcomes.

[0060] When resource availability parameters indicate limited bilingual resources or inadequate technical infrastructure, the system adopts a bilingual approach to core concepts. Under this strategy, the system prioritizes bilingual teaching of core scientific concepts, while secondary content can be taught solely in the local language or English. This core concept bilingual approach concentrates limited resources on ensuring the quality of instruction for key content under limited resource conditions, guaranteeing students master core knowledge points and acquire basic bilingual scientific vocabulary.

[0061] When local teaching context parameters indicate significant differences in language proficiency among students, the system employs a tiered bilingual teaching strategy. Under this strategy, students are divided into three levels of language proficiency: high, medium, and low. Different bilingual ratios and content difficulty levels are allocated to students at each level. Students with high language proficiency receive primarily English-language instruction, students with medium proficiency receive balanced bilingual instruction, and students with low proficiency receive primarily local language instruction. This tiered bilingual teaching strategy allows the system to adapt to the heterogeneity of the student population, providing each student with teaching content appropriate to their level.

[0062] Teacher support module 5 provides strategy adjustment instructions to learner model building engine 1, multilingual resource matching module 2, and learning path optimization module 3. Each module adjusts its working parameters according to these instructions. For example, when using a core concept bilingual combination strategy, multilingual resource matching module 2 prioritizes resources marked as core knowledge points for bilingual configuration, and does not need to adjust the language ratio for non-core knowledge point resources. When using a tiered bilingual teaching strategy, learner model building engine 1 additionally marks the student's language proficiency level when generating a two-dimensional learner feature profile, and multilingual resource matching module 2 configures different language ratios according to the level markings.

[0063] Through the flexible configuration of Teacher Support Module 5, the system can adapt to the actual teaching conditions and student characteristics of different regions and schools, providing teachers with operable teaching support tools and enhancing the system's practicality and scalability.

[0064] The data flow process of this invention system is as follows: Learner model building engine 1 collects students' learning behavior data and assessment result data, generates a two-dimensional learner feature profile, and passes it to multilingual resource matching module 2 and teacher support module 5. Multilingual resource matching module 2 extracts bilingual science education resources from the resource library based on the two-dimensional learner feature profile, applies a language ability adaptive weighting strategy to generate a language adaptive resource set, and passes it to learning path optimization module 3. Learning path optimization module 3 establishes a dual-objective optimization model based on the language adaptive resource set, solves for the optimal learning sequence, generates a personalized bilingual learning path, and passes it to content difficulty adjustment module 4 and teacher support module 5. Content difficulty adjustment module 4 monitors students' learning progress data in real time. When it detects a deviation from the expected progress threshold, it sends a difficulty adjustment instruction to multilingual resource matching module 2 and updates the language adaptive resource set. Teacher support module 5 receives local teaching context parameters and resource availability parameters, configures corresponding teaching strategies, and provides strategy adjustment instructions to each module. The modules achieve efficient data sharing and collaborative work through a data bus, forming a closed-loop adaptive system.

[0065] In a complete teaching scenario, a primary school has 30 third-grade students: 12 local students with high language proficiency; 10 new immigrant students with moderate language proficiency; and 8 recent immigrant students with low language proficiency. The school is equipped with basic multimedia teaching facilities and possesses some bilingual science education resources. Teacher support module 5, considering the significant differences in language proficiency among the students and the limited resources, is configured with a hybrid strategy combining differentiated bilingual teaching and a core concept bilingual integration strategy.

[0066] For the 12 students with high language proficiency, the system is configured with a learning path that uses 65%–75% English. Bilingual scientific terminology is presented simultaneously when explaining core concepts, while English is primarily used for non-core content. For the 10 students with intermediate language proficiency, the system is configured with a balanced bilingual learning path that uses 45%–55% English. All core concepts are explained bilingually, while non-core content is flexibly configured based on resource availability. For the 8 students with low language proficiency, the system is configured with a learning path that uses 70%–80% native language. Core concepts are explained in detail in the native language, supplemented with English terminology, while the native language is primarily used for non-core content.

[0067] After one semester of implementation, the class's average science test score improved from 65 to 86, a 32% increase. Students with low language proficiency saw the most significant improvement, rising from 55 to 78, a 42% increase. The class's average English scientific vocabulary increased by 128 words, and grammatical accuracy improved by an average of 38%. Teachers reported that the system effectively addressed the challenge of varying language proficiency levels among students, enabling students from different language backgrounds to develop both scientific knowledge and language skills in a learning environment suited to their individual levels.

[0068] This invention can also be implemented in several other ways. For example, the learner model building engine can introduce more dimensional evaluation indicators, such as students' learning motivation and learning style preferences, to form a more comprehensive learner profile. The multilingual resource matching module can support configurations of three or more languages, adapting to more complex multilingual teaching scenarios. The learning path optimization module can introduce more constraints, such as course progress requirements and exam schedules, to generate learning paths that better meet actual teaching needs. The content difficulty adjustment module can adopt a predictive adjustment strategy, predicting potential future difficulties based on students' historical learning data and adjusting the content difficulty in advance to achieve a smoother learning experience. The teacher support module can provide more refined strategy configuration options, allowing teachers to personalize the process according to specific teaching objectives and class characteristics.

[0069] It should be noted that, in this specification, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. The terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0070] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. An adaptive bilingual science education content intelligent recommendation system, characterized in that: include: A learner model building engine is used to collect students' learning behavior data and assessment result data. Based on the learning behavior data and the assessment result data, a scientific concept comprehension model and a language proficiency model are built respectively. By analyzing the interaction between the scientific concept comprehension model and the language proficiency model, a two-dimensional learner feature profile is generated. A multilingual resource matching module, connected to the learner model building engine, is used to receive the dual-dimensional learner feature profile, extract bilingual science education resources from the resource library based on the language proficiency parameters and scientific concept comprehension parameters in the dual-dimensional learner feature profile, determine the display ratio of English content and local language content in the bilingual science education resources based on the language ability adaptive weighting strategy, and generate a language adaptive resource set. The learning path optimization module, connected to the multilingual resource matching module, is used to receive the language adaptive resource set, establish a scientific knowledge acquisition objective function and a language ability development objective function, and determine the optimal learning sequence by solving a dual-objective optimization problem under the conditions of learning time constraints and cognitive load constraints, thereby generating a personalized bilingual learning path. The content difficulty adjustment module is connected to the learning path optimization module and the multilingual resource matching module. It is used to monitor students' learning progress data in real time. When the learning progress data is detected to deviate from the expected progress threshold, the difficulty level of scientific concepts and language expression of the bilingual science education resources is dynamically adjusted, and the language adaptive resource set is updated. The teacher support module, connected to the learner model building engine, the multilingual resource matching module, and the learning path optimization module, is used to receive local teaching context parameters and resource availability parameters, configure teaching strategies based on the local teaching context parameters and resource availability parameters, and provide strategy adjustment instructions to the learner model building engine, the multilingual resource matching module, and the learning path optimization module.

2. The adaptive bilingual science education content intelligent recommendation system according to claim 1, characterized in that: The learner model building engine includes a scientific concept comprehension assessment unit and a language proficiency assessment unit. The scientific concept comprehension assessment unit determines scientific concept comprehension parameters based on students' correct answer rate in science tests, the completeness of concept map drawing, and experimental operation proficiency. The language proficiency assessment unit determines language proficiency parameters based on students' vocabulary mastery, grammatical accuracy, and semantic comprehension ability.

3. The adaptive bilingual science education content intelligent recommendation system according to claim 1, characterized in that: The learner model building engine also includes an interaction impact analysis unit, which identifies the degree of impact of language barriers on science learning by calculating the correlation coefficient between the scientific concept comprehension parameter and the language proficiency parameter, and generates language support demand indicators.

4. The adaptive bilingual science education content intelligent recommendation system according to claim 1, characterized in that: The language proficiency adaptive weighting strategy in the multilingual resource matching module includes: when the language proficiency parameter is lower than a first preset threshold, increasing the display ratio of local language content to 70%–85% and decreasing the display ratio of English content to 15%–30%; when the language proficiency parameter is higher than a second preset threshold, increasing the display ratio of English content to 60%–80% and decreasing the display ratio of local language content to 20%–40%; and when the language proficiency parameter is between the first preset threshold and the second preset threshold, maintaining a balanced display ratio between English content and local language content.

5. The adaptive bilingual science education content intelligent recommendation system according to claim 1, characterized in that: The multilingual resource matching module also includes a bilingual scientific terminology comparison generation unit. This unit is used to identify core scientific terms in the bilingual science education resources, generate a comparison between the English and local language expressions of the core scientific terms, and simultaneously present the bilingual scientific terms during resource display.

6. The adaptive bilingual science education content intelligent recommendation system according to claim 1, characterized in that: The learning path optimization module uses a multi-objective genetic algorithm to solve the bi-objective optimization problem. The scientific knowledge acquisition objective function is established based on the knowledge point coverage and mastery level, and the language ability development objective function is established based on the vocabulary increment and grammatical complexity increment. The optimal learning sequence is determined through the Pareto optimal solution set.

7. The adaptive bilingual science education content intelligent recommendation system according to claim 6, characterized in that: The learning duration constraint requires that the duration of a single learning unit not exceed 30 minutes, and the cognitive load constraint requires that there be no more than two consecutive high-difficulty content units in the learning sequence.

8. The adaptive bilingual science education content intelligent recommendation system according to claim 1, characterized in that: The content difficulty adjustment module includes a progress monitoring unit and a difficulty adjustment unit. The progress monitoring unit collects students' learning completion time, answer accuracy rate and frequency of asking for help in real time. By comparing the deviation of the learning completion time from the expected time and the deviation of the answer accuracy rate from the expected accuracy rate, it determines whether there is a deviation from the expected progress threshold. The difficulty adjustment unit reduces or increases the difficulty level of the scientific concept and the difficulty level of the language expression according to the direction of the deviation.

9. The adaptive bilingual science education content intelligent recommendation system according to claim 1, characterized in that: The teaching strategies configured in the teacher support module include: when the resource availability parameter indicates that bilingual resources are sufficient, a full bilingual immersion teaching strategy is adopted; when the resource availability parameter indicates that bilingual resources are limited, a core concept bilingual combination strategy is adopted; and when the local teaching context parameter indicates that there are significant differences in the language proficiency of the student group, a tiered bilingual teaching strategy is adopted.

10. The adaptive bilingual science education content intelligent recommendation system according to claim 1, characterized in that: The learner model building engine, the multilingual resource matching module, the learning path optimization module, the content difficulty adjustment module, and the teacher support module are connected through a data bus to achieve data sharing and collaborative work.

Images