CN-121235120-B - Target language learning and developing system based on staged immersive context generation

Abstract

The invention discloses a target language learning and developing system based on staged immersive context generation, which comprises a learning data acquisition module, a cognition diagnosis module, a stage context resource module, an interactive training and feedback module, a stage updating module, a migration control and path scheduling module, a learning resource set, a multi-mode interactive training and feedback module, a stage updating module and a migration control and path scheduling module, wherein the learning data acquisition module is used for acquiring multi-source data and processing the multi-source data and outputting a standardized learning data set, the cognition diagnosis module is used for outputting a staged capability label based on a layered Bayesian cognition diagnosis model, the stage context resource module is used for calling the learning resource set according to the staged capability label, the interactive training and feedback module is used for carrying out multi-mode interactive training, acquiring performance data in real time and generating error correction information and minimization prompts, the stage updating module is used for executing posterior probability updating of potential variables and generating new staged capability labels and generating migration signals, and the migration control and path scheduling module is used for receiving the migration signals and controlling the learning path scheduling. The hierarchical Bayesian diagnosis method realizes the immersive learning path optimization of the target language.

Inventors

• GAO LIZE

Assignees

• 青岛蓝翼天擎教育科技有限公司

Dates

Publication Date

20260508

Application Date

20251015

Claims (5)

1. 1. A target language learning and development system based on staged immersive context generation, comprising: the learning data acquisition module is used for acquiring and processing multi-source data and outputting a standardized learning data set; the cognitive diagnosis module is used for receiving the standard learning data set, outputting a stage performance label based on the hierarchical Bayesian cognitive diagnosis model; the cognitive diagnostic module specifically includes: Inputting the standardized learning data set into a hierarchical Bayes cognitive diagnosis model, and establishing a potential variable layer and an observation layer in the hierarchical Bayes cognitive diagnosis model; Setting a hierarchical dependency structure in a potential variable layer, defining vocabulary potential variables and grammar potential variables as low-level potential variables, defining hearing potential variables and spoken potential variables as middle-level potential variables, and defining reading potential variables and writing potential variables as high-level potential variables; The prior condition of the middle latent variable is defined to depend on the posterior result of the lower latent variable, and the prior condition of the higher latent variable is defined to depend on the posterior result of the middle latent variable; Setting prior distribution for each potential variable, wherein each potential variable defines initial distribution through a group of super parameters; Defining likelihood functions between an observation layer and a potential variable layer, modeling each observation feature in the standardized learning data set one by one, and forming the overall probability of generating all observation data of a learner in a potential variable state; Performing posterior probability calculation on each potential variable through a Bayes inference formula to obtain a potential variable posterior probability set, and sequentially updating posterior distributions of middle-layer potential variables and high-layer potential variables based on a hierarchical dependency structure; outputting the mastering probability of the learner based on the posterior probability of each potential variable after updating, and generating a stage performance label according to a preset threshold; the stage context resource module is used for storing and managing the context learning resources divided according to the stages, and calling a learning resource set matched with the current stage of the learner according to the stage performance capability label; the interactive training and feedback module is used for carrying out multi-mode interactive training, collecting the performance data in real time and generating error correction information and minimized prompt; the interactive training and feedback module specifically comprises: selecting context learning resources corresponding to the capability labels of the learner at the current stage in the combined learning resource set, and constructing an immersive interaction scene; Performing interactive training of six dimensions of vocabulary, grammar, hearing, reading, spoken language and writing in the immersive interactive scene; collecting real-time performance data of learners in the immersive interactive training process, and constructing a real-time performance data set; Inputting the real-time performance data set into a voice recognition module, a natural language understanding module and a voice evaluation module to generate an error correction information set; determining a minimum prompt required by learner input based on the error correction information set, namely selecting one prompt which is most critical to the current input from all error correction information as an output prompt, and generating a minimum prompt result; The phase updating module is used for receiving real-time performance data, executing potential variable posterior probability updating, generating a new phase performance label and generating a migration signal when the advanced condition is met; The stage updating module specifically comprises: Inputting real-time performance data into a hierarchical Bayesian cognitive diagnosis model, and updating a potential variable layer by utilizing the real-time performance data based on a hierarchical dependency structure, so as to keep the corresponding relation of vocabulary potential variables, grammar potential variables, hearing potential variables, reading potential variables, spoken potential variables and writing potential variables; Under the condition that a real-time performance data set is obtained by executing posterior probability update on each potential variable, carrying out joint calculation by utilizing likelihood functions of the real-time performance data in the potential variable state and prior distribution of the original potential variable, and solving the posterior probability of the potential variable through a Bayesian inference formula, so as to obtain mastery probability of a learner in each capability dimension under the condition of the real-time performance data; In the hierarchical dependency structure, the posterior probability of the low-level potential variable is used as the prior input of the middle-level potential variable, the posterior probability of the middle-level potential variable is used as the prior input of the high-level potential variable, and the hierarchical updating from bottom to top is sequentially completed; generating a new stage capability label according to a preset threshold mapping rule according to the updated potential variable posterior probability; Judging the advanced stage condition of each dimension stage capability label, if the advanced stage condition is met, generating a migration signal and triggering stage migration, otherwise, maintaining the original stage state unchanged; The migration control and path scheduling module is used for receiving migration signals, driving the stage context resource module to call learning resources of a higher stage, controlling the learning path scheduling and recording the stage change of a learner.
2. 2. A target language learning and development system based on staged immersive context generation according to claim 1, wherein the modules are implemented by the following method: the method comprises the steps of collecting and processing multi-source data of a learner in a language training process to form a standard learning data set; inputting the standard learning data set into a hierarchical Bayes cognitive diagnosis model, and outputting a stage performance label; According to the stage performance labels, screening learning resources matched with the current stage of a learner from a stage context resource library to obtain a combined learning resource set; carrying out multi-mode interactive training on a learner in an immersive context, collecting real-time performance data of the learner in the interactive training process, and generating error correction information and a minimized prompt; Inputting the real-time performance data set into a hierarchical Bayesian cognitive diagnosis model, executing potential variable posterior probability update, generating a new stage performance label, and generating a migration signal when the advanced condition is met; based on the new stage capability label, stage context resource screening, immersive context interaction training and stage updating are re-executed until stage migration is triggered, so that a dynamic closed-loop language learning path is formed.
3. 3. The system of claim 2, wherein the standardized learning data set includes a pre-processed vocabulary reaction time, a vocabulary accuracy rate, a grammar structure recognition accuracy rate, a grammar error number, a hearing keyword recognition rate, a hearing answer delay time, a reading finish reading accuracy rate, a reading overreading accuracy rate, a spoken voice recognition similarity, a spoken pronunciation score, a writing grammar accuracy rate, and a writing vocabulary diversity index.
4. 4. The system for learning and developing a target language based on staged immersive context generation according to claim 2, wherein the step of selecting learning resources matched with the current stage of the learner from the staged context resource library according to the staged capability tag to obtain a combined learning resource set specifically comprises: Establishing a stage context resource library, dividing learning resources according to indexes of different stages, wherein each stage index set corresponds to one group of learning resources; correspondingly matching the stage capability label with a stage index set in a stage context resource library; In the matching process, mapping the stage capability labels and the stage index sets in the stage context resource library correspondingly to obtain learning resource sets of each dimension; and combining the learning resource sets obtained by screening the six-dimensional stage performance labels respectively to form a combined learning resource set.
5. 5. The system for learning and developing a target language based on staged immersive context generation according to claim 2, wherein the language learning path for forming a dynamic closed loop by re-executing staged context resource screening, immersive context interaction training and staged updating based on the new staged capability label until stage migration is triggered comprises: based on the new stage capability labels, invoking a stage context resource library, and determining matched learning resource sets corresponding to the stage capability labels of all dimensions to form a new combined learning resource set; Inputting the new combined learning resource set into the immersive context interactive training and real-time feedback step to generate a new real-time expression data set; Inputting the new real-time performance data set into a hierarchical Bayesian cognitive diagnosis model, calculating the posterior probability of the updated potential variable, and generating a new stage performance label; Judging whether the new stage capability label meets the advanced condition, if so, generating a migration signal and triggering stage migration; if the learning data is not satisfied, continuously screening the learning resources based on the new stage capability label, generating a new combined learning resource set, developing new immersive context interaction training, collecting new real-time performance data, inputting the hierarchical Bayesian cognitive diagnosis model again for updating, and forming a language learning path of a dynamic closed loop.

Description

Target language learning and developing system based on staged immersive context generation Technical Field The invention relates to the technical field of intelligent language learning, in particular to a target language learning and cultivating system based on staged immersive context generation. Background The existing target language learning system generally adopts a training mode of fixed courses and unified difficulty, and generally provides contents of vocabulary, grammar, hearing, reading, spoken language and writing through an online platform or mobile application. Although such methods can cover basic skills, they lack dynamic recognition of individual differences of learners, and are difficult to effectively adjust according to the cognitive stage of learners. Some systems attempt to introduce a self-adaptive mechanism, but most of the systems take the answer accuracy as a basis, and cannot comprehensively model multidimensional features, so that the personalized effect is limited. In the prior art, there are attempts to analyze learning behaviors by using a statistical model or a machine learning method, but the method lacks a clear hierarchical structure in stage division and advanced judgment, and the output result is relatively one-sided and cannot reflect the dependency relationship among different language skills. Meanwhile, learning resource recommendation is based on manual rules or shallow feature matching, resource screening and context generation cannot form a closed loop with a stage judgment result, and the learning process is lack of continuity. In addition, the feedback link of the existing system is mostly simple prompt or direct error correction, and cannot be combined with multi-source data to realize refined analysis, so that the problem of excessive or insufficient prompt content exists. For migration judgment in the learning stage, mathematics and repeatable judgment standards are also commonly lacking, experience thresholds are more depended, and objectivity is difficult to ensure. Thus, how to provide a target language learning and development system based on staged immersive context generation is a problem that needs to be solved by those skilled in the art. Disclosure of Invention The invention aims to provide a target language learning and cultivating system based on staged immersive context generation, which utilizes multi-source learning data acquisition processing, hierarchical Bayesian inference, context learning resource screening and multi-mode interactive training feedback to describe a closed-loop mechanism for driving learning resource screening, interactive training and stage updating through staged capability labels in detail, and has the advantages of high individuation degree, dynamic adjustability of learning paths and improvement of language learning and obtaining efficiency. A target language learning and development system based on staged immersive context generation according to an embodiment of the present invention includes: the learning data acquisition module is used for acquiring and processing multi-source data and outputting a standardized learning data set; the cognitive diagnosis module is used for receiving the standard learning data set, outputting a stage performance label based on the hierarchical Bayesian cognitive diagnosis model; the stage context resource module is used for storing and managing the context learning resources divided according to the stages, and calling a learning resource set matched with the current stage of the learner according to the stage performance capability label; the interactive training and feedback module is used for carrying out multi-mode interactive training, collecting the performance data in real time and generating error correction information and minimized prompt; The phase updating module is used for receiving real-time performance data, executing potential variable posterior probability updating, generating a new phase performance label and generating a migration signal when the advanced condition is met; The migration control and path scheduling module is used for receiving migration signals, driving the stage context resource module to call learning resources of a higher stage, controlling the learning path scheduling and recording the stage change of a learner. Optionally, the modules are realized by the following method: the method comprises the steps of collecting and processing multi-source data of a learner in a language training process to form a standard learning data set; inputting the standard learning data set into a hierarchical Bayes cognitive diagnosis model, and outputting a stage performance label; According to the stage performance labels, screening learning resources matched with the current stage of a learner from a stage context resource library to obtain a combined learning resource set; carrying out multi-mode interactive training on a learner in an immersive context, collect