Language barrier assessment method and system

Abstract

The invention discloses a language disorder assessment method and system, and belongs to the technical field of data processing, and the method comprises the steps that a user side accesses a language assessment platform, a language cognition task is generated, the language cognition task comprises a main language task and a secondary cognition task, and task sequence generation is carried out with cognition load increase; issuing the language cognition task to a user side thread, executing a task language test, and returning a task language flow; and a cognitive evaluation module embedded in the language evaluation platform is triggered, cascade evaluation based on logic paradox, language entropy change and a high-dimensional semantic network is executed, a language evaluation result is generated, and popup window display is performed on a platform user side. According to the method and the device, the technical problem that potential or slight language barriers are difficult to effectively induce and detect due to single language evaluation environment and lack of cognitive pressure in the prior art is solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, in particular to a language disorder evaluation method and system. BACKGROUND

[0002] The precise evaluation of language disorders is an important topic in clinical speech pathology and neurocognitive science. With the development of computer technology, although there are some auxiliary tools based on speech recognition or simple language feature analysis, the evaluation dimensions are often single, and are mostly limited to lexical and grammatical surface features, which are difficult to effectively capture the deep disorder patterns of language and thinking interaction under cognitive load, resulting in insufficient recognition sensitivity to early or mild disorders.

[0003] The existing technical solutions are relatively static in simulating cognitive load, and lack of design of dynamic increase of task difficulty. At the same time, the analysis method is often isolated, and multiple dimensions such as logical consistency, language information entropy and semantic network structure cannot be effectively fused and cascaded, resulting in limited recognition ability of complex language disorder patterns, and there is an urgent need for a more comprehensive evaluation method and system. SUMMARY

[0004] The present application provides a language disorder evaluation method and system, aiming to solve the technical problems of single language evaluation environment and lack of cognitive pressure in the prior art, which makes it difficult to effectively induce and detect potential or mild language disorders.

[0005] In view of the above problems, the present application provides a language disorder evaluation method and system.

[0006] The first aspect of the present application provides a language disorder evaluation method, which comprises: a user terminal accesses a language evaluation platform and generates a language cognitive task, wherein the language cognitive task contains a primary language task and a secondary cognitive task, and a task sequence is generated in the direction of cognitive load increase; the language cognitive task is downloaded to the user terminal thread, the task language test is executed, and the task language stream is returned; a cognitive evaluation module embedded in the language evaluation platform is triggered to perform cascaded evaluation based on logical paradox, language entropy change and high-dimensional semantic network, and the language evaluation result is generated and displayed on the platform user terminal pop-up window.

[0007] Another aspect of this application discloses a language barrier assessment system, which includes: a language cognitive task generation module for user terminal access to a language assessment platform to generate language cognitive tasks, wherein the language cognitive tasks include primary language tasks and secondary cognitive tasks, and the task sequence is generated in the direction of increasing cognitive load; a task language stream feedback module for sending the language cognitive tasks to a user terminal thread, executing task language tests, and feeding back the task language stream; and a language assessment result generation module for triggering the cognitive assessment module embedded in the language assessment platform to perform cascaded assessments based on logical paradoxes, language entropy changes, and high-dimensional semantic networks, generating language assessment results and displaying them in a pop-up window on the platform's user terminal.

[0008] One or more technical solutions provided in this application have at least the following technical effects or advantages: By adopting a user-end access platform and generating a composite paradigm that includes primary language tasks and secondary cognitive tasks, and arranging the task sequence in ascending order of cognitive load, this technical solution solves the technical problems of existing technologies, such as a single assessment environment and lack of cognitive pressure, which makes it difficult to effectively induce and detect potential or mild language disorders. It achieves the technical effect of significantly improving the ecological validity of the assessment, sensitively capturing functional defects that can only be manifested under the competition of cognitive resources, and providing high-quality, high-information time-series data for subsequent refined analysis.

[0009] The above description is merely an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, specific embodiments of this application are given below. Attached Figure Description

[0010] Figure 1 A flowchart illustrating the language barrier assessment method is provided for embodiments of this application.

[0011] Figure 2 A schematic diagram of the structure of a language barrier assessment system is provided for an embodiment of this application.

[0012] Figure labeling: Language cognition task generation module 11, task language flow feedback module 12, language assessment result generation module 13. Detailed Implementation

[0013] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided below.

[0014] The overall concept of the technical solution provided in this application is as follows: This application provides a method and system for assessing language barriers. Through interaction between a user terminal and an assessment platform, a complex task sequence is generated. This sequence combines primary language tasks with secondary cognitive tasks, increasing in difficulty according to cognitive load. This design aims to simulate the multitasking pressure of real-world communication, thereby more sensitively eliciting and assessing potential language barriers that are difficult to detect under simple, static testing.

[0015] After introducing the basic principles of this application, various non-limiting embodiments of this application will be described in detail below with reference to the accompanying drawings.

[0016] Example 1, as Figure 1 As shown in the embodiments of this application, a language barrier assessment method is provided, the method comprising: Step S100: The user terminal accesses the language assessment platform and generates a language cognition task, wherein the language cognition task includes a primary language task and a secondary cognitive task, and the task sequence is generated in the direction of increasing cognitive load.

[0017] Specifically, the language assessment platform is responsible for task management, data analysis, and result generation, and users access the platform via the internet. Language cognitive tasks refer to interactive tasks designed to assess a user's language ability. They are not single tasks but rather a composite task structure. The main language task is the core part of the task, directly targeting the language function to be assessed. Examples include story retelling, picture description, and word fluency tests. Secondary cognitive tasks are performed concurrently with the main task, aiming to utilize additional cognitive resources such as working memory and attention. For example, when completing the main task, users are required to tap their fingers in sync to a specific rhythm. Cognitive load progression refers to the design of the task sequence difficulty in a direction that gradually increases cognitive load. That is, subsequent tasks require more mental resources than preceding tasks, thus gradually putting pressure on the user's cognitive system.

[0018] Specifically, the user client connects to the language assessment platform's backend service via network protocols. The language assessment platform is built using a layered microservice architecture. Specifically, the backend core uses frameworks such as Spring Cloud to develop service modules including task management, data acquisition, and the assessment engine. The assessment engine integrates three core cognitive pathways: a logic paradox analysis module based on the Drools rule engine, an entropy calculation module based on Apache Spark MLlib, and a high-dimensional network construction module based on topology data analysis libraries such as GUDHI. The platform connects to the frontend application through an API gateway, uses Kafka message queues to process high-concurrency language stream data, and relies on time-series databases such as InfluxDB to store timestamped task data. Finally, it is deployed in the cloud via Docker containerization to achieve elastic scaling and modular iteration. After receiving a connection, the platform, based on preset assessment goals and user profiles such as age and suspected disability type, superimposes and pairs a primary language task with one or more secondary cognitive tasks, and quantifies the total cognitive load level based on task complexity, real-time requirements, and resource contention. Following the principle of "increasing" workload from low to high, these tasks are sorted and assembled into a complete sequence of tasks with progressively increasing difficulty.

[0019] For example, assessing executive function and language fluency involves the following main language task: Word retrieval – Name as many animals as possible. Secondary cognitive task: Working memory interference – Simultaneously, remember randomly flashing numbers on the screen and repeat them after the language task. Cognitive load increasing sequence: Sequence 1, low load: Only perform the animal name recitation task. Sequence 2, medium load: Perform the animal name recitation task while a number slowly flashes in the corner of the screen (mild interference). Sequence 3, high load: Perform the animal name recitation task while remembering three rapidly flashing different numbers.

[0020] This step, by introducing secondary cognitive tasks and cognitive load escalation sequences, effectively simulates the multitasking and psychological stress scenarios required for real-world dialogue in a controlled environment. This provides a high-quality, information-rich time-series data foundation for subsequent cascade analysis.

[0021] Step S200: Send the language cognition task to the user-end thread, perform the task language test, and send back the task language stream.

[0022] Specifically, a client-side thread refers to an independent execution sequence within a client-side application. To avoid UI lag, task execution typically occurs in a separate background thread, separate from the main thread displaying the user interface. The task language stream refers to the initially encapsulated data packets generated when the user performs a task. It includes not only the raw audio stream but also synchronously contains behavioral data related to secondary tasks, and all data is accurately timestamped.

[0023] Specifically, the language evaluation platform sends serialized task instruction packets to connected client machines via push messaging services such as WebSocket or MQTT. Upon receiving the instruction, the client machine creates an independent worker thread to ensure task execution does not block the user interface. Audio recording begins in this thread, and the corresponding sensor listeners are activated. After the task is completed, timestamped data, including audio, interaction events, and task metadata, is serialized into a structured data packet, transmitted to the language evaluation platform, and stored in temporary storage or a message queue for further processing.

[0024] This step ensures smooth user interaction through multi-threaded isolation and utilizes high-precision timestamp synchronization technology to achieve strict alignment of primary and secondary task multimodal data in the time dimension, thereby constructing a high-quality task language flow.

[0025] Step S300: Trigger the cognitive assessment module embedded in the language assessment platform to perform a cascaded assessment based on logical paradox, language entropy change and high-dimensional semantic network, generate language assessment results and display them in a pop-up window on the platform user end.

[0026] Specifically, cascaded evaluation is a phased, conditionally triggered evaluation process. In this application, it specifically refers to first performing a rapid first-order evaluation, and then automatically triggering a more complex second-order evaluation if the result is unclear.

[0027] Specifically, upon receiving the complete task language stream, the process follows a pre-defined cascading procedure: First-order evaluation of the language stream is performed by invoking logical paradoxes and language entropy changes within the first evaluation area. Judgments are made based on a rule set. If either result exceeds a pre-defined cognitive impairment threshold, it is considered a clear indication, and the result fusion unit is instructed to weightedly fuse the two results to directly generate the final evaluation result. If both results are below the threshold, the second evaluation area is triggered, initiating a high-dimensional semantic network analysis service for second-order evaluation, and the resulting topological indicators are incorporated as key evidence into the final result. Finally, the generator formats the data into a JSON or XML data packet and sends it to the user terminal in real-time via a WebSocket connection or HTTP push technology. Upon receiving the data packet, the user terminal invokes its UI components to render a modal pop-up window at the top layer of the interface, displaying the evaluation result to the user in a graphical and textual format.

[0028] This step, through cascading evaluation and real-time result feedback, integrates the entire system into a highly efficient automated diagnostic closed loop. It balances evaluation efficiency and depth. Furthermore, the task language test is performed, and the task language stream is returned, including: receiving a first language stream based on the primary language task and a second data stream based on the secondary cognitive task from the target user; and performing time-series synchronization timestamps on the first language stream and the second data stream as the task language stream, wherein the task language stream is a time-series data stream based on cognitive load increase.

[0029] Specifically, the first language stream refers to the raw data generated by the target user when performing the primary language task, typically an audio stream. For example, continuous speech signals recorded by a microphone when a user retells a story or describes a picture. The second data stream refers to the behavioral data stream generated by the target user when performing secondary cognitive tasks. It takes various forms, such as tapping coordinate sequences on a touchscreen, key press sequences, eye movement trajectory data, or motion signals from an accelerometer.

[0030] Specifically, based on the multimedia and sensor frameworks provided by the operating system on the user's device, an audio recording interface is initiated to capture the first speech stream, and corresponding sensor event listeners are registered to capture the second data stream. The system clock is invoked to timestamp the moment of acquisition commencement and on each subsequent data packet or event. For example, the audio stream is divided into consecutive frames at a fixed sampling rate, and each frame is marked with a start timestamp; simultaneously, each user touch of the screen is treated as an independent event and immediately marked with the precise timestamp of the event occurrence. Subsequently, the data with a unified time reference are arranged or associated chronologically and packaged into a complete, structured data file, such as encapsulating the audio file path and a list of timestamped events in JSON format, thus forming the task speech stream.

[0031] This step uses high-precision timestamp identification technology to achieve strict alignment of heterogeneous data from different sources and with different sampling rates in the time dimension, transforming the original audio and behavioral signals into a task-oriented language stream with temporal relationships.

[0032] Furthermore, before triggering the cognitive assessment module embedded in the language assessment platform, the construction of the first cognitive branch includes: setting a semantic logic symbol pattern, constructing a word conversion threshold using the semantic logic symbol pattern; using the semantic logic symbols as the data pattern, constructing a semantic inference engine through adversarial learning based on logical paradox; and coupling the word conversion threshold with the semantic inference engine as the first cognitive branch.

[0033] Specifically, semantic logic symbol patterns refer to predefined rules used to map words, phrases, and syntactic relations in natural language into formalized logical symbols. A term conversion threshold is a filtering or judgment criterion used to determine whether a word or phrase is important enough to be converted into a logical symbol, and how to convert it. For example, setting a frequency threshold to convert only content words that appear more than N times, or a confidence threshold to convert only named entity recognition confidence values ​​higher than X. A semantic inference engine is a program or model capable of automatically reasoning about symbolized logical expressions. It can check consistency, derive implicit conclusions, etc. The First Cognitive Branch is a complete analysis module coupled with term conversion thresholds and semantic inference engines; its function is to evaluate the logical consistency and reasoning ability of the user's language.

[0034] Specifically, a semantic logic symbolic pattern is defined, which involves creating a domain-specific dictionary and grammatical rules. Based on this pattern, a term conversion threshold is constructed. Specifically, user language flow is processed through word segmentation and part-of-speech tagging, and key entities and their relationships are extracted using named entity recognition and dependency parsing. According to preset frequency or importance thresholds, standard terms and syntactic structures are converted into logical symbols. Using these symbols as data patterns, a semantic inference engine is constructed based on symbolic logic and adversarial learning frameworks. Specifically, formalized semantic axioms and inference rules are defined using first-order logic or descriptive logic, and training is performed using generative adversarial networks or reinforcement learning. The generator is responsible for constructing interfering proposition sequences containing logical paradoxes, such as self-contradictions and circular references, while the discriminator, i.e., the inference engine ontology, performs logical consistency checks and implication relation judgments on the input symbol sequences through a Prolog kernel. During adversarial training, the generator continuously challenges the discriminator's inference boundaries, enabling it to learn to identify more subtle logical errors, thereby improving its sensitivity and robustness to abnormal logical structures. The output is a logic verification model capable of handling symbolic propositions. Finally, the output of the term conversion threshold is directly fed to the semantic inference engine through a software interface call, thus completing the coupling between the two and forming the first cognitive branch.

[0035] This step allows the assessment to go beyond the correctness of vocabulary or the standardization of grammar, and to delve deeper into the underlying cognitive barriers that users face when using language to reason, argue, and maintain narrative coherence. It is applicable to assessing language impairments related to executive function and reasoning ability.

[0036] Furthermore, before triggering the cognitive assessment module embedded in the language assessment platform, the construction of the second cognitive branch includes: setting a language entropy dimension, wherein the language entropy dimension includes at least lexical entropy, syntactic entropy and information entropy; and generating the second cognitive branch by constructing a language-cognitive load collapse curve based on the entropy value calculation of the language entropy dimension and the coupling of multiple entropies.

[0037] Specifically, language entropy refers to different computational levels used to quantify the uncertainty and confusion of language. Lexical entropy measures the richness and diversity of vocabulary. It typically calculates the probability distribution of word occurrences within a given window of words; a higher entropy value indicates a richer and more varied vocabulary, while a lower entropy value may suggest repetitive or stereotypical vocabulary. Syntactic entropy measures the complexity and variability of syntactic structures. By analyzing the probability distribution of sentence structures, such as using context-free grammars, a high entropy value indicates flexible and complex sentence structures, while a low entropy value indicates simple and straightforward sentence structures. Information entropy, originating from information theory, measures the uncertainty or amount of information conveyed by language. It assesses the difficulty of predicting the next language unit given a context. A high entropy value indicates high information density or difficulty in prediction.

[0038] Specifically, for lexical entropy, NLTK and other tool libraries are used for word segmentation and word frequency statistics, and the entropy is calculated based on the Shannon entropy formula. For syntactic entropy, a parser, such as Stanford Parser or spaCy dependency parsing, is used to convert sentences into syntax trees, and the entropy value is calculated based on the distribution changes of the tree structure. For information entropy, an N-gram language model is used to calculate the perplexity or conditional entropy of word sequences. This involves preprocessing a large-scale corpus, including word segmentation and cleaning; calculating the joint probability of word sequences using statistical methods; sliding a window of length N in the training text to count the frequency of all N-tuples; calculating the conditional probability of each word given the context of the previous N-1 words based on maximum likelihood estimation; and employing smoothing techniques to assign probabilities to out-of-vocabulary word tuples to generate a probability lookup table. The output of this table measures the statistical characteristics and uncertainty of the language sequence and forms the basis for calculating information entropy. The system receives multiple language streams generated by users under different cognitive load levels, calculates the lexical entropy, syntactic entropy, and information entropy of each language stream in parallel, and obtains a comprehensive entropy value representing the current language state through a feature fusion algorithm. Simultaneously, using task difficulty as the X-axis and the corresponding comprehensive entropy value as the Y-axis, a language-cognitive load collapse curve is fitted using a nonlinear regression method. This curve is the final output of the second cognitive branch.

[0039] This step, by introducing multi-dimensional entropy calculation, transforms subjective language fluency, richness, and coherence into quantifiable objective indicators. This provides robust data support for the assessment, enabling assessors not only to determine the existence of obstacles but also to quantify the severity and specific manifestations of those obstacles, significantly improving the accuracy and interpretability of the assessment.

[0040] Furthermore, before triggering the cognitive assessment module embedded in the language assessment platform, the construction of the third cognitive branch includes: deploying a network reconstruction layer based on the semantic geometric network reconstruction of the user's language flow, wherein the reconstruction target is the thought pattern structure; constructing a high-dimensional manifold semantic network, and constructing the third cognitive branch by mapping the semantic geometric network in the high-dimensional manifold semantic network and calculating topological invariants, wherein the topological invariants at least include a first evaluation array based on the Betti number and persistent cohomology, and a second evaluation array based on the Lyapunov index and the association dimension.

[0041] Specifically, semantic geometric network reconstruction refers to representing a user's speech flow as a network graph. Nodes represent key concepts or words, and edges represent semantic or grammatical relationships between words. Mathematically, this network can be viewed as a geometric object. A high-dimensional manifold semantic network is a pre-constructed background semantic space containing a large number of concepts and complex relationships. It can be understood as embedding all possible words and relationships into a high-dimensional mathematical space, whose shape is curved and complex, i.e., a manifold. Topological invariants are mathematical quantities used to describe the property of a geometric shape to remain unchanged under continuous deformation, such as stretching or bending, without tearing or sticking together. They focus on the overall connectivity of the shape rather than specific details. Betti numbers are used to describe the number of holes of different dimensions in a geometric shape. For example, the 0th Betti number represents the number of connected components, and B1 represents the number of circular holes. In semantic networks, a high B1 value may suggest the presence of circular arguments or self-referential structures in the thought process. Persistent cohomology refers to calculating the lifespan of topological features at different scales. Short-lived holes may be noise, while long-lived holes represent stable and essential structural features of the network. The Lyapunov index, derived from dynamical systems theory, measures a system's sensitivity to initial conditions. In semantic networks, it is used to quantify the unpredictability or chaos of thought leaps.

[0042] Specifically, the network reconstruction layer utilizes dependency parsing and co-occurrence statistics to transform a user's speech stream into a concrete semantic geometric network. This semantic geometric network is then mapped onto a pre-trained high-dimensional manifold semantic network, which serves as a reference frame. This high-dimensional manifold semantic network is typically trained on a large-scale corpus using word embedding models such as Word2Vec, GloVe, or BERT. Specifically, using a large-scale corpus, words are mapped to high-dimensional vectors through word embedding models such as Word2Vec, GloVe, or BERT, with each vector representing a point of a word in the semantic space. Based on the cosine similarity or transformed geometric distance between words, a weighted network is constructed, where nodes are word vectors and edge weights reflect the strength of semantic association. To further characterize the overall geometric structure of this vector space, manifold learning algorithms, such as UMAP, t-SNE, or autoencoders, are introduced to reduce the dimensionality of the high-dimensional vectors and visualize them, thus approximating a low-dimensional manifold representation. Finally, a hybrid model combining discrete graph structure and continuous geometric representation is constructed, providing a foundation for topology analysis. Next, using topological data analysis tools, the topological structure of the user network after being mapped to a high-dimensional manifold is analyzed, and the first evaluation array is calculated, including the Betti numbers of different dimensions and the persistence of holes in each dimension, i.e., their lifetime. Simultaneously, by analyzing the evolution of concept nodes on the manifold over time or in narrative order, their dynamical system characteristics are calculated, resulting in the second evaluation array, including Lyapunov exponents and correlation dimensions. Finally, the complete analytical pipeline described above, capable of outputting complex topologies and dynamical system invariants, constitutes the third cognitive branch.

[0043] In this step, the third cognitive approach achieves a paradigm shift in language assessment by introducing topological data analysis and dynamic systems methods; it delves deeper from analyzing the surface features of language to quantifying its inherent cognitive geometry and dynamic characteristics. It can extremely sensitively capture subtle cognitive structural disturbances that traditional methods cannot detect, such as fragmented thinking, closed-loop thinking, or cognitive chaos. This makes this approach particularly suitable for screening early, mild cognitive impairments and identifying complex cognitive disorders in the mental health field.

[0044] Furthermore, triggering the cognitive assessment module embedded in the language assessment platform includes: writing the first cognitive branch and the second cognitive branch into the first assessment area in parallel, and writing the third cognitive branch into the second assessment area to form the cognitive assessment module; performing a first-order assessment of the task language flow based on the first assessment area, and performing a second-order assessment based on the second assessment area, wherein the second assessment area can be selectively triggered.

[0045] Specifically, the first evaluation area is a logical container dedicated to handling first-order evaluation. It is designed to run the first and second cognitive branches in parallel. These two branches are computationally lightweight and fast, designed for initial screening. The second evaluation area is a separate container or module that handles second-order evaluation. It is dedicated to running the third cognitive branch. This branch is computationally complex and time-consuming, used for in-depth, detailed analysis. First-order evaluation refers to the rapid, preliminary analysis performed by the first evaluation area. Its goal is to quickly determine whether there are obvious or typical obstacle features in the language flow using computationally efficient methods. Second-order evaluation refers to the in-depth, secondary analysis performed by the second evaluation area. It is triggered when the results of the first-order evaluation are unclear, aiming to discover hidden, atypical, or early obstacle features. Optional triggering means that the activation of the second evaluation area is not default but controlled by a predefined decision rule. Second-order evaluation is automatically triggered only when the results of the first-order evaluation meet specific conditions; otherwise, it is skipped to save resources.

[0046] Specifically, the first and second cognitive branches are packaged into two independent Docker containers, receiving tasks via message queues to achieve parallel processing. These two containers together constitute the first evaluation area. Similarly, the third cognitive branch is deployed as another independent container, serving as the second evaluation area. The entire process is controlled by a workflow decision engine. When the task language stream arrives, the engine starts the two branches within the first evaluation area in parallel. Simultaneously, a decision service continuously monitors the output results of these two branches and compares them with preset cognitive impairment thresholds. Optional triggering logic: if either the first or second branch's result exceeds the threshold, the decision service determines a diagnosis, immediately generates a result, terminates the process, and skips the second evaluation area; only when both results are below the threshold will the decision service send an instruction to the workflow engine to start the container in the second evaluation area for computation. This step, by constructing partitioned and cascaded assessment modules and implementing an optional triggering mechanism, achieves the core technical effect of optimizing system computational efficiency while ensuring assessment depth and sensitivity. Separating rapid screening from precise diagnosis forms a highly efficient analysis pipeline: for typical cases, the system can respond quickly through a lightweight first-order assessment fast channel; only for atypical or mild cases is the deep detection channel activated. This makes the technical solution particularly suitable for large-scale population screening, clinical environments with high real-time requirements, or edge devices with limited computing resources, achieving a balance between accurate assessment and practicality.

[0047] Furthermore, a cascaded evaluation based on logical paradox, language entropy change, and high-dimensional semantic networks is performed, including: according to the first cognitive branch, semantic logical symbol transformation is performed on the task language flow, and the first cognitive result is determined by logical reasoning of symbol elements; according to the second cognitive branch, multidimensional entropy value is calculated on the task language flow, and the horizontal axis is constructed with the cognitive load increase direction, and the vertical axis is constructed with the superposition state entropy value, to construct the second cognitive result based on the language-cognitive load collapse curve.

[0048] Specifically, the first cognitive outcome is the output of the first cognitive branch, which is usually a quantified score or indicator used to represent the degree of impairment in the user's language in terms of logical consistency and reasoning ability.

[0049] Specifically, when the task language flow enters the first evaluation area, two branches are activated simultaneously: In the first cognitive branch, its semantic-logical symbol converter first converts the user's language into a sequence of symbols. Subsequently, the semantic inference engine performs reasoning analysis on these symbols, detecting logical errors. Based on the number, type, and severity of the detected errors, the first cognitive result is determined. In the second cognitive branch, multidimensional entropy calculation is performed to obtain lexical entropy, syntactic entropy, and information entropy, and these entropy values ​​are combined into a superposition entropy value. Using the difficulty level of the multiple task sequences completed by the user as the horizontal axis and the superposition entropy value corresponding to each task as the vertical axis, a language-cognitive load collapse curve is plotted to obtain the second cognitive result.

[0050] This step achieves a multi-faceted and objective first-order rapid assessment of language barriers by performing two different dimensions of quantitative analysis in parallel. The first cognitive pathway explores the depth of logical connotation, effectively revealing deficiencies in thinking coherence and reasoning ability; while the second cognitive pathway measures from the breadth of information theory, sensitively capturing the degradation patterns of language fluency and complexity under cognitive stress.

[0051] Furthermore, the first cognitive result and the second cognitive result are verified; if at least one of the first cognitive result and the second cognitive result is higher than the cognitive impairment threshold, the first cognitive result and the second cognitive result are fused to generate a language assessment result.

[0052] Specifically, the cognitive impairment threshold is one or more critical scores pre-trained using large amounts of data from both normal and disabled individuals. It is used to distinguish between normal and abnormal ranges. For example, the logical contradiction score threshold is set to 0.7, and the entropy change collapse index threshold is set to 0.6.

[0053] Specifically, the cognitive impairment threshold is learned from clinical data. The system receives first and second cognitive results, calculated in parallel from the first assessment area. These results are compared and judged. If at least one of the first and second cognitive results exceeds the cognitive impairment threshold, a weighted average algorithm is used. This algorithm assigns weights based on the prediction accuracy of each branch, mapping the two scores to the final language assessment result. For example, the rule might be: if the logic score is high and the entropy change score is normal, it is judged as reasoning impairment; if both are high, it is judged as mixed severe impairment. Example illustration. This step, through the establishment of a threshold-based verification mechanism and result fusion, enables efficient decision-making and conclusion generation in the cascading evaluation process.

[0054] Furthermore, if both the first cognitive result and the second cognitive result are less than the cognitive impairment threshold, the third cognitive branch in the second evaluation area is activated and input into the task language stream; by performing semantic geometric network reconstruction on the task language stream, matching is performed in the high-dimensional manifold semantic network, parallel evaluation and coupling based on the first evaluation array and the second evaluation array are performed, the third cognitive result is determined and added to the language evaluation result.

[0055] Specifically, when both the first and second cognitive results are below a threshold, the second evaluation area is activated. The third cognitive branch, based on dependency parsing tools such as spaCy and co-occurrence statistics, constructs a semantic geometric network from the concepts and relations extracted from the task's language flow. This network is then mapped onto a high-dimensional manifold semantic network. After mapping, the topology analysis module, typically calling a TDA library such as GUDHI, calculates the topological features of the network in the high-dimensional space and outputs the first evaluation array; simultaneously, the dynamical system analysis module treats the sequence of concept nodes appearing on the manifold as a dynamical system and calculates its second evaluation array. The results of these two arrays are coupled to generate the third cognitive result, which is added as key evidence to the final language evaluation result.

[0056] This step enables the system to detect the earliest, mildest, or most atypical cognitive impairments that have not yet affected outward language expression but have led to simplification, fragmentation, or chaos in thought structure. This greatly reduces the false negative rate and provides an objective technical means for ultra-early warning, accurate differential diagnosis, and tracking of intervention effects.

[0057] In summary, the language barrier assessment method provided in this application has the following technical effects: 1. By constructing a composite paradigm of primary language tasks and secondary cognitive tasks, and a task sequence with increasing cognitive load, the ecological validity of the assessment scenario was improved. It can effectively simulate the multitasking pressure in real communication, thereby sensitively inducing mild or potential language impairments that are difficult to manifest in a single-task calm state.

[0058] 2. By synchronizing and aligning multimodal data streams using high-precision timestamps, it becomes possible to analyze the real-time interaction between primary and secondary tasks under cognitive load, enhancing the accuracy and objectivity of the assessment and providing key data support for understanding the competition for cognitive resources.

[0059] 3. By constructing partitioned and cascaded assessment modules and implementing an optional triggering mechanism, the system's computational efficiency is optimized while ensuring assessment depth and sensitivity. Rapid screening and precise diagnosis are separated to form an efficient pipeline: rapid response to typical cases, and in-depth analysis only initiated for atypical cases. This makes the solution suitable for large-scale screening and real-time application scenarios.

[0060] Example 2, based on the same inventive concept as the language barrier assessment method in the foregoing examples, such as... Figure 2 As shown in the embodiment of this application, a language barrier assessment system is provided. The system includes: a language cognitive task generation module 11, used for user terminal access to a language assessment platform to generate language cognitive tasks, wherein the language cognitive tasks include primary language tasks and secondary cognitive tasks, and the task sequence is generated in the direction of cognitive load increase; a task language flow feedback module 12, used for sending the language cognitive tasks to the user terminal thread, executing task language tests, and feeding back the task language flow; and a language assessment result generation module 13, used for triggering the cognitive assessment module embedded in the language assessment platform, executing cascaded assessment based on logical paradox, language entropy change, and high-dimensional semantic network, generating language assessment results, and displaying them in a pop-up window on the platform user terminal.

[0061] Furthermore, the task language stream feedback module 12 is also used to perform the following steps: receiving a first language stream based on the primary language task and a second data stream based on the secondary cognitive task from the target user; performing time-series synchronization timestamp identification on the first language stream and the second data stream as the task language stream, wherein the task language stream is a time-series data stream based on cognitive load increase.

[0062] Furthermore, the system is also used to perform the following steps: setting a semantic logic symbol pattern, constructing a term conversion threshold using the semantic logic symbol pattern; constructing a semantic inference engine using the semantic logic symbol as a data pattern and through adversarial learning based on logical paradox; and coupling the term conversion threshold with the semantic inference engine as a first cognitive branch.

[0063] Furthermore, the system is also used to perform the following steps: setting a language entropy dimension, wherein the language entropy dimension includes at least lexical entropy, syntactic entropy and information entropy; and generating a second cognitive branch by calculating the language-cognitive load collapse curve under multi-entropy coupling based on the entropy value of the language entropy dimension.

[0064] Furthermore, the system is also used to perform the following steps: reconstructing a semantic geometric network based on user language flow, deploying a network reconstruction layer, wherein the reconstruction target is the thought pattern structure; constructing a high-dimensional manifold semantic network, and constructing a third cognitive branch by mapping the semantic geometric network in the high-dimensional manifold semantic network and calculating topological invariants, wherein the topological invariants at least include a first evaluation array based on Betti number and persistence cohomology, and a second evaluation array based on Lyapunov index and association dimension.

[0065] Furthermore, the language evaluation result generation module 13 is also used to perform the following steps: writing the first cognitive branch and the second cognitive branch into the first evaluation area in parallel, and writing the third cognitive branch into the second evaluation area to form the cognitive evaluation module; performing a first-order evaluation of the task language flow according to the first evaluation area, and performing a second-order evaluation according to the second evaluation area, wherein the second evaluation area can be selectively triggered.

[0066] Furthermore, the language evaluation result generation module 13 is also used to perform the following steps: according to the first cognitive branch, perform semantic logical symbol conversion on the task language stream, and determine the first cognitive result by performing logical reasoning of symbol elements; according to the second cognitive branch, perform multidimensional entropy calculation on the task language stream, construct the horizontal axis with the cognitive load increase direction, construct the vertical axis with the superposition state entropy value, and construct the second cognitive result based on the language-cognitive load collapse curve.

[0067] Furthermore, the language assessment result generation module 13 is also used to perform the following steps: verifying the first cognitive result and the second cognitive result; if at least one of the first cognitive result and the second cognitive result is higher than the cognitive impairment threshold, fusing the first cognitive result and the second cognitive result to generate a language assessment result.

[0068] Furthermore, the language evaluation result generation module 13 is also used to perform the following steps: if both the first cognitive result and the second cognitive result are less than the cognitive impairment threshold, activate the third cognitive branch in the second evaluation area and input it into the task language stream; reconstruct the semantic geometric network of the task language stream, perform matching in the high-dimensional manifold semantic network, perform parallel evaluation and coupling based on the first evaluation array and the second evaluation array, determine the third cognitive result and add it to the language evaluation result.

[0069] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any modifications, equivalent changes, and alterations 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 scope of the present invention.

Claims

1. A language barrier assessment method, characterized in that, The method includes: The user accesses the language assessment platform and generates language cognitive tasks, which include primary language tasks and secondary cognitive tasks, and the task sequence is generated in order to increase cognitive load. The language recognition task is sent to the user-end thread to perform a language test and then the language stream is returned. The cognitive assessment module embedded in the language assessment platform is triggered to perform a cascaded assessment based on logical paradox, language entropy change, and high-dimensional semantic network, generating language assessment results and displaying them in a pop-up window on the platform user end.

2. The language barrier assessment method as described in claim 1, characterized in that, Perform the task language test and return the task language stream, including: Receive the target user's first language stream based on the primary language task and the second data stream based on the secondary cognitive task; The first language stream and the second data stream are time-series synchronized with a timestamp identifier to form the task language stream, wherein the task language stream is a time-series data stream based on cognitive load increase.

3. The language barrier assessment method as described in claim 1, characterized in that, Before triggering the cognitive assessment module embedded in the language assessment platform, the construction of the first cognitive pathway includes: Define a semantic logic symbol pattern, and construct a term conversion threshold based on the semantic logic symbol pattern; Using semantic logic symbols as data patterns, a semantic reasoning machine is constructed through adversarial learning based on logical paradoxes. The term conversion threshold is coupled to the semantic reasoning machine as the first cognitive branch.

4. The language barrier assessment method as described in claim 3, characterized in that, Before triggering the cognitive assessment module embedded in the language assessment platform, the construction of the second cognitive pathway includes: Define a language entropy dimension, wherein the language entropy dimension includes at least lexical entropy, syntactic entropy and information entropy; A second cognitive branch is generated by constructing a language-cognitive load collapse curve based on the entropy value calculation based on the language entropy dimension and the coupling of multiple entropies.

5. The language barrier assessment method as described in claim 4, characterized in that, Before triggering the cognitive assessment module embedded in the language assessment platform, the construction of the third cognitive pathway includes: A network reconstruction layer is deployed based on semantic geometric network reconstruction of user language flow, with the reconstruction target being the thought pattern structure; A high-dimensional manifold semantic network is constructed, and a third cognitive branch is constructed by mapping the semantic geometric network in the high-dimensional manifold semantic network and calculating the topological invariants. The topological invariants include at least a first evaluation array based on the Betti number and persistence cohomology, and a second evaluation array based on the Lyapunov index and the association dimension.

6. The language barrier assessment method as described in claim 5, characterized in that, Triggering the cognitive assessment module embedded in the language assessment platform includes: The first cognitive branch and the second cognitive branch are written into the first evaluation area in parallel, and the third cognitive branch is written into the second evaluation area to form the cognitive evaluation module. The task language stream is evaluated in a first-order manner based on the first evaluation region, and in a second-order manner based on the second evaluation region, wherein the second evaluation region is optionally triggered.

7. The language barrier assessment method as described in claim 6, characterized in that, Perform cascaded evaluations based on logical paradoxes, language entropy changes, and high-dimensional semantic networks, including: Based on the first cognitive branch, the task language stream is subjected to semantic-logical symbol transformation, and the first cognitive result is determined by logical reasoning of the symbol elements. Based on the second cognitive branch, multidimensional entropy values ​​are calculated for the task language flow. The horizontal axis is constructed with the cognitive load increase direction, and the vertical axis is constructed with the superposition state entropy value, thus constructing a second cognitive result based on the language-cognitive load collapse curve.

8. The language barrier assessment method as described in claim 7, characterized in that, Verify the first cognitive result and the second cognitive result; If at least one of the first cognitive results and the second cognitive results is higher than the cognitive impairment threshold, the first cognitive result and the second cognitive result are merged to generate a language assessment result.

9. The language barrier assessment method as described in claim 8, characterized in that, If both the first cognitive result and the second cognitive result are less than the cognitive impairment threshold, the third cognitive branch in the second assessment area is activated and input into the task language stream; By reconstructing the semantic geometry network of the task language stream, matching is performed in a high-dimensional manifold semantic network, parallel evaluation and coupling based on the first evaluation array and the second evaluation array are executed, and a third cognitive result is determined and added to the language evaluation result.

10. A language barrier assessment system, characterized in that, The system for performing the language impairment assessment method according to any one of claims 1 to 9, the system comprising: The language cognition task generation module is used for user terminal access to the language assessment platform to generate language cognition tasks. The language cognition tasks include primary language tasks and secondary cognitive tasks, and the task sequence is generated in the direction of increasing cognitive load. The task language stream feedback module is used to send the language cognition task to the user-end thread, perform task language testing, and feedback the task language stream. The language assessment result generation module is used to trigger the cognitive assessment module embedded in the language assessment platform, perform cascaded assessments based on logical paradoxes, language entropy changes, and high-dimensional semantic networks, generate language assessment results, and display them in a pop-up window on the platform user end.

Images