A method and system for testing vocabulary comprehension combining spelling and pronunciation

Abstract

The application discloses a vocabulary comprehensive ability test method and system combining spelling and pronunciation, and relates to the technical field of data processing. The method comprises the following steps: collecting spelling data, voice data and correction operation data of a user on target vocabulary; calculating bimodal mastery values of each target vocabulary based on the spelling data and the voice data, and calculating correction cost values of each target vocabulary based on the correction operation data; mapping each target vocabulary to a dynamic cognitive state matrix, and marking vocabulary with a bimodal mastery value lower than a first threshold value and a correction cost value higher than a second threshold value as high-risk vocabulary; if the bimodal mastery values of more than a preset proportion of associated vocabulary of the high-risk vocabulary decrease or the correction cost values of the associated vocabulary increase, it is determined that there is a vocabulary network instability risk formed by the high-risk vocabulary and the associated vocabulary; and generating a diagnosis report according to the dynamic cognitive state matrix and the determination result of the vocabulary network instability risk.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, specifically to a method and system for testing comprehensive vocabulary ability by combining spelling and pronunciation. Background Technology

[0002] In the process of language learning, vocabulary mastery is the core foundation, and the ability is reflected in the comprehensive cognition of the form, sound, and meaning of words. Traditional vocabulary testing methods mostly focus on single-modal assessment, such as written spelling tests or independent oral pronunciation tests. Although these methods can examine spelling accuracy or pronunciation standardization separately, they have obvious limitations. They sever the inherent connection between "form" and "sound" in the mental lexicon and cannot assess the depth of learners' mastery of the rules of word sound-form correspondence. Moreover, they rely solely on the correctness of the final answer for judgment, ignoring the learners' cognitive processing behaviors during the answering process (such as hesitation and modification), while these behavioral data contain key information about the stability of vocabulary representation and cognitive load.

[0003] Some existing computerized testing systems have begun to attempt to record reaction time or simple error types, but their analytical logic remains relatively superficial, typically stopping at scoring and categorizing isolated words. They fail to examine learners' errors within their entire vocabulary cognitive network, and therefore cannot diagnose deeper, systemic cognitive weaknesses, such as the pervasive absence of a particular spelling rule across multiple related words, or an unstable grasp of a word family as a whole. Summary of the Invention

[0004] The purpose of this invention is to provide a method and system for testing comprehensive vocabulary ability by combining spelling and pronunciation, so as to solve the problems raised in the prior art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a method for testing comprehensive vocabulary ability by combining spelling and pronunciation, the method comprising: S100. Collect the user's spelling data, speech data, and correction operation data for the target word. The correction operation data includes the correction type, the time of correction, and the content of correction. The collection process continues until the user completes all test operations for the target word and submits the results. S200: Calculate the bimodal mastery value of each target word based on the spelling data and the speech data, and calculate the correction cost value of each target word based on the correction operation data. S300. Based on the bimodal mastery value and the correction cost value, each target word is mapped to a dynamic cognitive state matrix, and words in the dynamic cognitive state matrix whose bimodal mastery value is lower than the first threshold and whose correction cost value is higher than the second threshold are marked as high-risk words. S400. Based on the lexical association network, analyze the bimodal mastery value and correction cost value of the associated words of the high-risk words in the dynamic cognitive state matrix. The lexical association network includes word root, semantic or phonological similarity associations between words. S500. If, among the related words of the high-risk word, more than a preset proportion of the related words have a decreased bimodal mastery value or an increased correction cost value, it is determined that there is a risk of instability in the word network composed of the high-risk word and its related words. S600. Generate a diagnostic report based on the dynamic cognitive state matrix and the determination result of the risk of instability of the vocabulary network.

[0006] According to the above scheme, the spelling data includes the complete character sequence of the user's spelling input for the target word, the total spelling input time, whether the spelling input is completed and submitted within the preset time limit, and the input order and corresponding input time of each character. The complete character sequence of the spelling input records the trajectory of all character changes from the first input to the final submission in real time, and the input time is accurate to the millisecond level. The voice data includes the user's pronunciation voice signal for the target word, the voice signal acquisition duration, the pronunciation start time, the pronunciation end time, and whether a valid pronunciation audio that meets the preset quality threshold is acquired. The voice signal acquisition meets the preset sampling rate and bit rate technical requirements. The quality threshold of the valid pronunciation audio is set based on a comprehensive consideration of background noise intensity, pronunciation clarity, and audio integrity. The correction type is the addition, deletion, or replacement of spelling characters, or the interruption and re-pronunciation of speech reading. The correction operation is triggered by the user actively triggering the editing command or by the input change recognized by the system. The correction occurrence time is the time interval between each operation and the start time of the test or the time of the previous operation. Continuous correction operations are recorded separately according to their respective trigger times. The correction content is recorded as the specific character and position information of the spelling character being edited, or the audio time period corresponding to the re-pronunciation of the speech reading. The position information is accurate to the index position of the character in the complete sequence, and the audio time period is accurate to the start frame and the end frame.

[0007] According to the above scheme, the dual-modal mastery values ​​include: S211. Compare the complete character sequence of the spelling input with the standard answer, and generate a spelling accuracy score through a character-by-character comparison algorithm; based on the total spelling input time, the input order of each character and the corresponding input time, generate a spelling fluency index by calculating the deviation of the variance of the input interval between characters from the average time. S212. Analyze the acoustic feature matching degree between the spoken speech signal and the standard speech model. Generate a speech accuracy score by calculating the dynamic time-warped distance of the Mel frequency cepstral coefficient sequence or the cosine similarity of the deep speech feature embedding vector. Based on the duration from the start of speech to the end of speech and the energy change characteristics of the spoken speech signal, generate a speech fluency index by analyzing the uniformity of energy packet distribution, the proportion of silent segments, and the smoothness of energy start and stop transitions within the speech segment. S213. Input the spelling accuracy score, the pronunciation accuracy score, the spelling fluency index and the pronunciation fluency index into the first fusion function. The first fusion function is a weighted linear combination function that presets the weight coefficients of each index according to the vocabulary assessment target. The weight coefficients are preset based on the importance ratio of spelling ability and pronunciation ability in the comprehensive assessment, and the sum of all weight coefficients is 1. Output the bimodal mastery value. The revised cost value includes: S221. Based on the correction operation data, count the total number of correction operations and identify the number of operations that belong to cross-modal correction. Cross-modal correction refers to the operation of correcting the pronunciation output based on the spelling input result or correcting the spelling input based on the pronunciation output result. S222. Based on the time of the correction, calculate the duration of each correction operation. The duration of a single correction operation is the time interval from the time of this correction trigger to the time of correction completion or the time of the next correction trigger. S223. Input the total number of correction operations, the number of cross-modal correction operations, and the duration of each correction operation into the second mapping function. The second mapping function is a function that performs nonlinear normalization processing according to a preset cost mapping rule. The cost mapping rule is preset based on the high cognitive load characteristics of cross-modal correction operations and the degree of influence of correction time on mastery. After normalization processing, the correction cost value ranges from 0 to 1. Output the correction cost value.

[0008] According to the above scheme, step S300 includes: S310. Using the bimodal mastery value as the first parameter and the modified cost value as the second parameter, construct a two-dimensional parameter space to characterize the cognitive state of each target word. S320. Each target word is determined as a state point in the two-dimensional parameter space based on its corresponding bimodal mastery value and correction cost value. The set of state points of all target words constitutes the dynamic cognitive state matrix. The dynamic cognitive state matrix is ​​indexed and stored according to the test order of the target words or the unique word identifier. S330. Set a first threshold and a second threshold in the two-dimensional parameter space; the first threshold and the second threshold are pre-calibrated based on the age characteristics, language proficiency level and difficulty level of the target test user group, and can be dynamically adjusted according to different test scenarios and evaluation objectives; the first threshold is the lowest qualified value of the bimodal mastery value, and the second threshold is the highest qualified value of the correction cost value. S340. Traverse all state points in the dynamic cognitive state matrix, mark target words that satisfy the condition that the bimodal mastery value is lower than the first threshold and the correction cost value is higher than the second threshold as high-risk words, and store all original test data and calculation indicators associated with the target words.

[0009] According to the above scheme, step S400 includes: S410. Construct a vocabulary association network, wherein the vocabulary association network is a graph structure with words as nodes and word root derivation relations, semantic similarity or phonetic similarity as edges. The data source for constructing the network is an authoritative dictionary, a large-scale corpus, and word formation and semantic association rules in vocabulary acquisition theory. The semantic similarity is calculated based on the cosine distance of words in the semantic vector space, the phonetic similarity is calculated based on the edit distance of phoneme sequences, and the word root derivation relations are determined based on word form rules. S420. For each marked high-risk word, according to the word association network, retrieve and determine at least one related word that is directly connected to it. Direct connection means that there is a single association edge between the two word nodes and no other intermediate word nodes. S430. Based on the dynamic cognitive state matrix, by querying the state point coordinates corresponding to each associated word, extract the bimodal mastery value and correction cost value of each associated word of the high-risk word, construct the associated word state set of the high-risk word, store the associated word state set according to the association relationship type, and associate it with the latest state of the dynamic cognitive state matrix in real time.

[0010] According to the above scheme, step S500 includes: S510. For each high-risk word, obtain its associated word status set; S520. Based on the dynamic cognitive state matrix, obtain the baseline bimodal mastery value and baseline correction cost value of the preset word family to which the high-risk words belong. The preset word family is pre-divided according to the rules of the same root, consistent semantic field or similar speech pattern. The baseline bimodal mastery value is the statistical mean of the bimodal mastery values ​​of all tested words in the word family. The baseline correction cost value is the statistical mean of the correction cost values ​​of all tested words in the word family. S530. In the set of associated word states, compare the bimodal mastery value of each associated word with the benchmark bimodal mastery value of its word family, count the first number of associated words whose bimodal mastery value is lower than the benchmark bimodal mastery value, compare the correction cost value of each associated word with the benchmark correction cost value of its word family, and count the second number of associated words whose correction cost value is higher than the benchmark correction cost value. S540. Perform aggregation calculation on the first quantity and the second quantity, wherein the aggregation calculation is to directly add the two quantities to obtain the number of risk words; calculate the ratio of the number of risk words to the total number of words in the associated word state set; if the ratio of the number of risk words to the total number of words in the associated word state set exceeds a preset ratio threshold, wherein the preset ratio threshold is preset based on the statistical analysis law of the stability of the word cognitive network, then it is determined that there is a risk of instability in the word network with the high-risk word as the core, and the determination result of the instability risk is associated with the identification and specific ratio data of the high-risk word and its associated words.

[0011] According to the above scheme, step S600 includes: S610. Extract high-risk words and their corresponding bimodal mastery values ​​and correction cost values ​​from the dynamic cognitive state matrix, and extract the list of related words and actual risk ratios of each unstable network from the judgment results of the unstable risk of the word network. S620. The actual risk ratio is compared with multiple preset risk level thresholds. The preset risk level thresholds are divided into at least three levels with a strict order according to the ratio range. Each level corresponds to a unique text or graphical risk level identifier, and the corresponding risk level identifier is determined. S630. According to the preset diagnostic report template and data field mapping relationship, organize and fill in the extracted high-risk vocabulary information, bimodal mastery value, corrected cost value, related vocabulary list, actual risk ratio and determined risk level identifier to generate a structured diagnostic report.

[0012] A comprehensive vocabulary ability testing system that combines spelling and pronunciation, comprising: a data acquisition module, a data analysis module, and a report generation module; The data acquisition module is used to collect the user's spelling data, pronunciation data, and correction operation data for the target words; the correction operation data includes the correction type, the time of correction, and the content of correction. The data analysis module is used to calculate the bimodal mastery value of each target word based on the spelling data and speech data, and to calculate the correction cost value of each target word based on the correction operation data; it is used to construct a dynamic cognitive state matrix based on the bimodal mastery value and the correction cost value, and to mark words with bimodal mastery values ​​below a first threshold and correction cost values ​​above a second threshold as high-risk words; it is used to analyze the state of associated words of the high-risk words based on the word association network to determine the risk of word network instability. The report generation module is used to generate a diagnostic report based on the dynamic cognitive state matrix and the determination result of the risk of instability of the vocabulary network.

[0013] According to the above scheme, the data analysis module includes a state assessment unit and a network analysis unit; The state assessment unit is used to calculate the bimodal mastery value and the correction cost value, construct the dynamic cognitive state matrix, and mark words in the dynamic cognitive state matrix whose bimodal mastery value is lower than the first threshold and whose correction cost value is higher than the second threshold as high-risk words. The network analysis unit is used to analyze the cognitive state of the associated words of the high-risk words based on the word association network, and to determine the risk of instability of the word network based on the deterioration ratio of the state of the associated words.

[0014] According to the above scheme, the report generation module includes a data integration unit and a report synthesis unit; The data integration unit is used to extract high-risk vocabulary information, quantitative indicators, and risk network composition data from the dynamic cognitive state matrix and network instability risk judgment results. The report synthesis unit is used to assess the risk level of the extracted data and generate a structured diagnostic report.

[0015] Compared with the prior art, the beneficial effects of the present invention are: 1. This invention integrates spelling and pronunciation data, combines accuracy and fluency multi-dimensional indicators to calculate the bimodal mastery value, and quantifies the cost and complexity of correction operations to accurately reflect the user's comprehensive vocabulary application ability and true cognitive state. 2. This invention constructs a dynamic cognitive state matrix based on bimodal mastery value and correction cost value, realizing a visual mapping of vocabulary cognitive state, and accurately identifying core weaknesses by combining high-risk vocabulary marking logic; 3. This invention constructs a lexical network by associating word roots, semantics, and phonetics, analyzes the risk transmission effect of high-risk words on related words, realizes the quantitative judgment of the risk of lexical network instability, and provides early warning of systemic errors. Attached Figure Description

[0016] Figure 1This is a flowchart illustrating the steps of a comprehensive vocabulary ability testing method that combines spelling and pronunciation according to the present invention. Figure 2 This is a schematic diagram of the structure of a vocabulary comprehensive ability testing system that combines spelling and pronunciation according to the present invention. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] Example: Figures 1-2 As shown, this invention provides a solution, a method for testing comprehensive vocabulary ability by combining spelling and pronunciation, the method comprising the following steps: S100. Collect the user's spelling data, speech data, and correction operation data for the target word. The correction operation data includes the correction type, the time of correction, and the content of correction. The collection process continues until the user completes all test operations for the target word and submits the results. Specifically, the spelling data includes the complete character sequence of the user's spelling input for the target word, the total spelling input time, whether the spelling input was completed and submitted within the preset time limit, and the input order and corresponding input time of each character. The complete character sequence of the spelling input records the trajectory of all character changes from the first input to the final submission in real time, with the input time accurate to the millisecond level. The voice data includes the user's pronunciation voice signal for the target word, the voice signal acquisition duration, the pronunciation start time, the pronunciation end time, and whether valid pronunciation audio that meets the preset quality threshold was acquired. The pronunciation voice signal acquisition meets the preset sampling rate and bit rate technical requirements, and the quality threshold of the valid pronunciation audio is based on the background. The noise level, pronunciation clarity, and audio integrity are comprehensively set. The correction type is the addition, deletion, or replacement of spelling characters or the interruption and re-pronunciation of speech. The correction operation is triggered by the user actively triggering the editing command or the input change recognized by the system. The correction time is the time interval between each operation and the start time of the test or the time of the previous operation. Continuous correction operations are recorded separately according to their respective trigger times. The correction content is recorded as the specific character and position information of the spelling character being edited, or the audio time period corresponding to the re-pronunciation of the speech. The position information is accurate to the index position of the character in the complete sequence, and the audio time period is accurate to the start frame and the end frame. For example: push the target words "analyze," "analysis," and "analytical" to the user in sequence; the user then completes the spelling and pronunciation test for each word in turn; Spelling data acquisition: Target word: analyze; User operation: Operation sequence: Input the letter 'a', pause briefly, input 'n', then input 'a' and 'l', then incorrectly input 'y', the user realizes the error, presses the backspace key to delete the 'y', then re-inputs the correct 'z', and finally inputs 'e' to complete the spelling; The system captures all the above keyboard events and their precise occurrence time in real time, and stores them as a sequence: (a,t=0), (n,t=150), (a,t=280), (l,t=410), (y,t=520), (Backspace,t=650), (z (e, t=780), (e, t=900); where t represents the number of milliseconds since the start of this spelling test; the system also records the total spelling time as T=900ms and confirms that the submission is completed within the preset time limit; the correction event is identified from the sequence, and the Backspace event that occurs at t=650ms is marked as a deletion type correction operation; its correction content is parsed as deleting the character y at index 4 (counting from 0, corresponding to the 5th character position); since this correction occurs after the user inputs y, it belongs to the spelling modality, and is triggered by the user's attempt to silently read and pronounce, i.e., the speech modality, so the system determines that this correction belongs to cross-modal correction.

[0019] Voice data acquisition: Target vocabulary analysis: After the user clicks the record button, the user reads the word "analysis" aloud; System recording: The system starts audio recording with a sampling rate of 16kHz and a depth of 16 bits; the recording continues from the moment the user begins to speak until the moment the speech ends, and the voice signal acquisition duration T is calculated. audio =1.2 seconds; The recorded audio is analyzed in real time. It is found that the background noise intensity is low, the pronunciation is clear, and the audio waveform is complete. Therefore, the audio segment is determined to be a valid pronunciation audio that meets the preset quality threshold. This is only an example and is not a limitation.

[0020] S200: Calculate the bimodal mastery value of each target word based on spelling data and speech data, and calculate the correction cost value of each target word based on correction operation data. Specifically, bimodal mastery values ​​include: S211. Compare the complete character sequence of the spell input with the standard answer, and generate a spelling accuracy score through a character-by-character comparison algorithm; based on the total spelling input time, the input order of each character and the corresponding input time, generate a spelling fluency index by calculating the deviation of the variance of the input interval between characters from the average time. For example: Compare the user's final submitted character sequence `analyze` with the standard answer `analyze`, using a character-by-character comparison algorithm; a perfect match is achieved, resulting in an accuracy score of A.spell =1.0; Calculate the time interval Δt between each adjacent character input. i For example, if the time intervals are 150ms, 130ms, 130ms, 110ms, 130ms, 130ms, and 120ms, calculate their mean μΔt and variance σΔt. 2 The smoothness index is F. spell =exp(-(σΔt 2 / (μΔt×k s ))), where k s The preset adjustment coefficient, F, indicates that the smaller the variance, the smoother the input. spell The closer it is to 1, the better the calculated F is. spell =0.92; This is just an example and is not a limitation. S212. Analyze the acoustic feature matching degree between the pronounced speech signal and the standard pronunciation model. Generate a pronunciation accuracy score by calculating the dynamic time-warped distance of the Mel frequency cepstral coefficient sequence or the cosine similarity of the deep speech feature embedding vector. Based on the duration from the start to the end of pronunciation and the energy change characteristics of the pronounced speech signal, generate a pronunciation fluency index by analyzing the uniformity of energy packet distribution, the proportion of silent segments, and the smoothness of energy start and stop transitions within the speech segment. For example: Extracting the MFCC feature sequence X from user analysis audio. user and the standard model MFCC sequence X std The minimum matching cost D is calculated using the dynamic time warping algorithm. DTW , used to represent the dynamic time-warped distance, measures the minimum cumulative difference between two MFCC feature sequences, and is formulated as: D DTW =min {π} Σ {(i,j)∈π} ||X user (i)-X std (j)|| 2 Where π represents the normalized path, and X is the mapping. user With X std A sequence of frame index pairs (i,j); X user (i) represents the MFCC feature vector of the i-th frame of the user's audio; X std (j) represents the MFCC feature vector of the j-th frame of the standard pronunciation model; The formula for scoring pronunciation accuracy is A. pron =exp(-α×D DTW ), where A pron The score represents the pronunciation accuracy, with values ​​closer to 1 indicating more accurate pronunciation; α represents the preset attenuation coefficient, and the dynamic time warping distance D... DTW The smaller the value, the higher the score; calculate A.pron =0.88; Analyze the energy envelope E(t) of the audio signal; calculate the energy variance Var(E) of the effective speech segment and the proportion of silence R. silence The formula for the fluency index is: F pron =1-(β1×Var(E)+β2×R silence ), where F pron The fluency index is represented by Var(E); Var(E) represents the variance of the energy envelope of the articulated segment, measuring energy stability; R silence This represents the proportion of silent segments in the total duration; β1 and β2 represent preset weighting coefficients used to balance the impact of energy variance and silent segment proportion on smoothness; F is calculated. pron =0.85; This is just an example and is not a limitation. S213. Input the spelling accuracy score, pronunciation accuracy score, spelling fluency index and pronunciation fluency index into the first fusion function. The first fusion function is a weighted linear combination function after preset weight coefficients for each index according to the vocabulary assessment goal. The weight coefficients are preset based on the importance ratio of spelling ability and pronunciation ability in the comprehensive assessment, and the sum of all weight coefficients is 1. Output the bimodal mastery value. For example: The first fusion function is a weighted linear combination: M = w1 × A spell +w2×F spell +w3×A pron +w4×F pron Where M represents the bimodal mastery value, a comprehensive quantitative indicator of vocabulary mastery; w1, w2, w3, and w4 represent the preset weights of the four indicators: spelling accuracy, spelling fluency, pronunciation accuracy, and pronunciation fluency, respectively, and satisfy w1+w2+w3+w4=1; preset weights: w1=0.3 (spelling accuracy), w2=0.2 (spelling fluency), w3=0.3 (pronunciation accuracy), w4=0.2 (pronunciation fluency); Substituting the data: M analyze =0.918; This is just an example and is not a limitation. The revised cost values ​​include: S221. Based on the correction operation data, count the total number of correction operations and identify the number of operations that are cross-modal corrections. Cross-modal corrections refer to operations that correct pronunciation output based on spelling input results, or correct spelling input based on pronunciation output results. For example: Count the number of operations: For the word "analyze", the total number of correction operations N total =1, where N is the number of cross-modal corrections. cross =1; N total N represents the total number of correction operations performed on a single word. cross This indicates the number of operations involved in the cross-modal correction. S222. Calculate the duration of each correction operation based on the time of occurrence. The duration of a single correction operation is the time interval from the time the current correction is triggered to the time the correction is completed or the time of the next correction trigger. For example, for a single correction operation, i.e., deleting y, the duration is ΔT. correct =780ms-650ms=130ms; ΔT correct Indicates the duration of a single correction operation; S223. Input the total number of correction operations, the number of cross-modal correction operations, and the duration of each correction operation into the second mapping function. The second mapping function is a function that performs non-linear normalization processing according to a preset cost mapping rule. The cost mapping rule is preset based on the high cognitive load characteristics of cross-modal correction operations and the degree of influence of correction time on mastery. After normalization processing, the value of the correction cost is in the range of 0 to 1. Output the correction cost value. For example: the second mapping function is: C raw =(N total +λ×N cross )×log(1+ΣΔT correct / T0); where C raw λ represents the original, unnormalized corrected cost value; λ represents the cross-modal correction penalty factor, where λ>1 indicates higher cross-modal correction costs; T0 represents the time normalization constant, used to adjust the scale of the impact of time on cost; log() represents the natural logarithm function; ΣΔT correct Represents the sum of the durations of all correction operations; calculate the original cost: C raw =0.3666; Here, λ=2, T0=1000ms; Nonlinear normalization to the [0,1] interval is performed using the sigmoid function: C=2 / (1+exp(-γ×C) raw ))-1; where C represents the normalized corrected cost value, with a value range of [0,1]. The higher the value, the greater the cognitive load; γ represents the preset scaling factor, used to control the steepness of the sigmoid function; exp() represents the natural exponential function; substituting into the calculation (taking γ=2): C=0.351; this is only an example and is not a limitation.

[0021] S300. Based on the bimodal mastery value and the correction cost value, each target word is mapped to the dynamic cognitive state matrix, and words in the dynamic cognitive state matrix whose bimodal mastery value is lower than the first threshold and whose correction cost value is higher than the second threshold are marked as high-risk words. Specifically, step S300 includes: S310. Using bimodal mastery value as the first parameter and modified cost value as the second parameter, construct a two-dimensional parameter space to characterize the cognitive state of each target word. S320. Each target word is determined as a state point in a two-dimensional parameter space based on its corresponding bimodal mastery value and correction cost value. The set of state points of all target words constitutes a dynamic cognitive state matrix. The dynamic cognitive state matrix is ​​indexed and stored according to the test order of the target words or the unique word identifier. For example, construct a two-dimensional parameter space and fill in the following values: analyze: M1=0.918, C1=0.351; analysis: M2=0.750, C2=0.620; analytical: M3=0.850, C3=0.200; where M represents the bimodal mastery value and C represents the corrected cost value. These values ​​form a set of state points: P1(0.918,0.351), P2(0.750,0.620), P3(0.850,0.200), which is the dynamic cognitive state matrix. This is only an example and is not a limitation. S330. Set a first threshold and a second threshold in a two-dimensional parameter space. The first threshold and the second threshold are pre-calibrated based on the age characteristics, language proficiency level and difficulty level of the target test user group, and can be dynamically adjusted according to different test scenarios and evaluation objectives. The first threshold is the lowest qualified value of the bimodal mastery value, and the second threshold is the highest qualified value of the correction cost value. S340. Traverse all state points in the dynamic cognitive state matrix, mark target words that satisfy the bimodal mastery value below the first threshold and the correction cost value above the second threshold as high-risk words, and store all original test data and calculation indicators associated with the target words. For example: when traversing a matrix, P2 (analysis) satisfies M2 = 0.750 < θ M =0.800 and C2=0.620>θ C =0.500, where θ M Let θ be the first threshold. C The second threshold is defined as target words whose bimodal mastery value is lower than the first threshold and whose correction cost value is higher than the second threshold. Therefore, "analysis" is marked as a high-risk word. S400. Based on the analysis of the vocabulary association network, the bimodal mastery value and correction cost value of the associated words of high-risk words in the dynamic cognitive state matrix are analyzed. The vocabulary association network includes the word root, semantic or phonological similarity association between words. Specifically, step S400 includes: S410. Construct a lexical association network. The lexical association network is a graph structure with words as nodes and word root derivation relations, semantic similarity or phonetic similarity as edges. The data sources for network construction are authoritative dictionaries, large-scale corpora and word formation and semantic association rules in vocabulary acquisition theory. Among them, semantic similarity is calculated based on the cosine distance of words in semantic vector space, phonetic similarity is calculated based on the edit distance of phoneme sequences, and word root derivation relations are determined based on word form rules. For example: Since analyze, analysis, and analytical share the root "analy-", according to word formation rules, a connection is established in the graph with these three words as nodes and root derivation as edges; S420. For each marked high-risk word, based on the word association network, retrieve and determine at least one related word that is directly connected to it. Direct connection means that there is a single association edge between the two word nodes and no other intermediate word nodes. For example, for the high-risk word analysis, retrieve its directly related words to get analyze and analytical. S430. Based on the dynamic cognitive state matrix, by querying the coordinates of the state points corresponding to each associated word, the bimodal mastery value and correction cost value of each associated word of the high-risk word are extracted, and a set of associated word states of the high-risk word is constructed. The set of associated word states is stored according to the type of association relationship and is linked to the latest state of the dynamic cognitive state matrix in real time. For example: Query from the dynamic cognitive state matrix: related word analyze: M1=0.918, C1=0.351; related word analytical: M3=0.850, C3=0.200; construct the related word state set of analysis: {(analyze: (0.918, 0.351)), (analytical: (0.850, 0.200))}; This is just an example and is not a limitation.

[0022] S500. If, among the related words of a high-risk word, more than a preset proportion of the related words have a decreased bimodal mastery value or an increased correction cost value, it is determined that there is a risk of instability in the word network composed of the high-risk word and its related words. Specifically, step S500 includes: S510. For each high-risk word, obtain its associated word status set; S520. Based on the dynamic cognitive state matrix, obtain the baseline bimodal mastery value and baseline correction cost value of the preset word family to which the high-risk words belong. The preset word family is pre-divided according to the rules of the same root, consistent semantic field or similar speech pattern. The baseline bimodal mastery value is the statistical mean of the bimodal mastery values ​​of all tested words in the word family. The baseline correction cost value is the statistical mean of the correction cost values ​​of all tested words in the word family. For example: Obtain the set of associated lexical states for "analysis"; these three words belong to the same predefined word family "analy-"; calculate the baseline value for this word family: the average bimodal mastery value M. bench =0.839; Average Corrected Cost C bench =0.390; where M bench This represents the baseline bimodal mastery value, i.e., the statistical mean of the M values ​​of all words within the word family; C bench This represents the baseline adjusted cost value, which is the statistical mean of the C values ​​of all words within the word family; S530. In the set of associated word states, compare the bimodal mastery value of each associated word with the baseline bimodal mastery value of its word family, count the first number of associated words whose bimodal mastery value is lower than the baseline bimodal mastery value, compare the modified cost value of each associated word with the baseline modified cost value of its word family, and count the second number of associated words whose modified cost value is higher than the baseline modified cost value. For example: comparing one by one: Analysis: M=0.918>M bench =0.839 (No); C=0.351 <C bench =0.390 (No); analytical: M=0.850>M bench =0.839 (No), C=0.200 <C bench =0.390 (No); Therefore, the first quantity is 0, and the second quantity is 0; S540. Perform aggregation calculation on the first quantity and the second quantity. Aggregation calculation is to directly add the two quantities to obtain the number of risk words. Calculate the ratio of the number of risk words to the total number of words in the associated word state set. If the ratio of the number of risk words to the total number of words in the associated word state set exceeds a preset ratio threshold, the preset ratio threshold is preset based on the statistical analysis law of the stability of the word cognitive network. Then, it is determined that there is a risk of instability in the word network with high-risk words as the core. The determination result of the instability risk is associated with the identification and specific ratio data of the high-risk words and their associated words. For example: Aggregate calculation: Number of risky words = 0 + 0 = 0; Total number of related words = 2; Risk ratio = 0 / 2 = 0.0; Preset ratio threshold is set to θrisk =0.5; θ risk This represents a preset ratio threshold used to determine whether the network is unstable; since 0.0 < 0.5, it is determined that there is no risk of network instability in the lexical network centered on analysis.

[0023] S600. Based on the dynamic cognitive state matrix and the judgment results of the risk of instability of the vocabulary network, generate a diagnostic report; Specifically, step S600 includes: S610. Extract high-risk words and their corresponding bimodal mastery values ​​and correction cost values ​​from the dynamic cognitive state matrix, and extract the list of related words and actual risk ratios of each unstable network from the judgment results of the unstable risk of the word network. S620. Compare the actual risk ratio with multiple preset risk level thresholds. The preset risk level thresholds are divided into at least three levels with a strict order according to the ratio range. Each level corresponds to a unique text or graphical risk level identifier, and the corresponding risk level identifier is determined. S630. According to the preset diagnostic report template and data field mapping relationship, organize and fill in the extracted high-risk vocabulary information, bimodal mastery value, corrected cost value, related vocabulary list, actual risk ratio and determined risk level identifier to generate a structured diagnostic report.

[0024] This invention provides another solution: a vocabulary comprehensive ability testing system that combines spelling and pronunciation, the system comprising: a data acquisition module, a data analysis module, and a report generation module; The data acquisition module is used to collect the user's spelling data, pronunciation data, and correction operation data for the target words; the correction operation data includes the correction type, the time of correction, and the content of the correction. The data analysis module is used to calculate the bimodal mastery value of each target word based on spelling and speech data, and to calculate the correction cost value of each target word based on correction operation data; it is used to construct a dynamic cognitive state matrix based on the bimodal mastery value and correction cost value, and to mark words with bimodal mastery values ​​below a first threshold and correction cost values ​​above a second threshold as high-risk words; it is used to analyze the status of related words of high-risk words based on the word association network to determine the risk of word network instability. Specifically, the data analysis module includes a state assessment unit and a network analysis unit. The state assessment unit is used to calculate the bimodal mastery value and the correction cost value, construct a dynamic cognitive state matrix, and mark words in the dynamic cognitive state matrix whose bimodal mastery value is lower than the first threshold and whose correction cost value is higher than the second threshold as high-risk words. The network analysis unit is used to analyze the cognitive state of the related words of high-risk words based on the word association network, and determine the risk of word network instability based on the state degradation ratio of the related words.

[0025] The report generation module is used to generate diagnostic reports based on the dynamic cognitive state matrix and the judgment results of the risk of instability in the vocabulary network; Specifically, the report generation module includes a data integration unit and a report synthesis unit. The data integration unit is used to extract high-risk vocabulary information, quantitative indicators, and risk network composition data from the dynamic cognitive state matrix and network instability risk assessment results. The report synthesis unit is used to assess the risk level of the extracted data and generate a structured diagnostic report.

[0026] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

Claims

1. A method for comprehensively testing vocabulary ability by combining spelling and pronunciation, characterized in that: The method includes: S100. Collect the user's spelling data, speech data, and correction operation data for the target word. The correction operation data includes the correction type, the time of correction, and the content of correction. S200: Calculate the bimodal mastery value of each target word based on the spelling data and the speech data, and calculate the correction cost value of each target word based on the correction operation data. S300. Based on the bimodal mastery value and the correction cost value, each target word is mapped to a dynamic cognitive state matrix, and words in the dynamic cognitive state matrix whose bimodal mastery value is lower than the first threshold and whose correction cost value is higher than the second threshold are marked as high-risk words. S400. Based on the lexical association network, analyze the bimodal mastery value and correction cost value of the associated words of the high-risk words in the dynamic cognitive state matrix. The lexical association network includes word root, semantic or phonological similarity associations between words. S500. If, among the related words of the high-risk word, more than a preset proportion of the related words have a decreased bimodal mastery value or an increased correction cost value, it is determined that there is a risk of instability in the word network composed of the high-risk word and its related words. S600. Generate a diagnostic report based on the dynamic cognitive state matrix and the determination result of the risk of instability of the vocabulary network.

2. The vocabulary comprehensive ability testing method combining spelling and pronunciation according to claim 1, characterized in that: The spelling data includes the complete character sequence of the user's spelling input for the target word, the total spelling input time, whether the spelling input was completed and submitted within the preset time limit, and the input order and corresponding input time of each character; The voice data includes the user's pronunciation voice signal for the target word, the duration of voice signal acquisition, the start time of pronunciation, the end time of pronunciation, and whether a valid pronunciation audio that meets the preset quality threshold has been acquired. The correction type is the addition, deletion, or replacement of spelling characters, or the interruption and re-pronunciation of speech reading; the correction time is the time interval between each operation and the start time of the test or the time of the previous operation; the correction content is recorded as the specific character and position information of the spelling characters being edited, or the audio time period corresponding to the re-pronunciation of the speech reading.

3. The vocabulary comprehensive ability testing method combining spelling and pronunciation according to claim 2, characterized in that: The bimodal mastery values ​​include: S211. Compare the complete character sequence of the spelling input with the standard answer to generate a spelling accuracy score; generate a spelling fluency index based on the total spelling input time, the input order of each character and the corresponding input time. S212. Analyze the acoustic feature matching degree between the spoken speech signal and the standard speech model to generate a speech accuracy score; generate a speech fluency index based on the duration from the start of speech to the end of speech and the energy change characteristics of the spoken speech signal. S213. Input the spelling accuracy score, the pronunciation accuracy score, the spelling fluency index and the pronunciation fluency index into the first fusion function, perform a weighted linear combination, and output the bimodal mastery value. The revised cost value includes: S221. Based on the correction operation data, count the total number of correction operations and identify the number of operations that belong to cross-modal correction. S222. Calculate the duration of each correction operation based on the time when the correction occurs; S223. Input the total number of correction operations, the number of cross-modal correction operations, and the duration of each correction operation into the second mapping function, perform nonlinear normalization processing, and output the correction cost value.

4. The vocabulary comprehensive ability testing method combining spelling and pronunciation according to claim 1, characterized in that: Step S300 includes: S310. Using the bimodal mastery value as the first parameter and the modified cost value as the second parameter, construct a two-dimensional parameter space to characterize the cognitive state of each target word. S320. Each target word is determined as a state point in the two-dimensional parameter space based on its corresponding bimodal mastery value and correction cost value. The set of state points of all target words constitutes the dynamic cognitive state matrix. S330. Set a first threshold and a second threshold in the two-dimensional parameter space; the first threshold is the lowest qualified value of the dual-modal mastery value, and the second threshold is the highest qualified value of the correction cost value. S340. Target words that meet the condition that the bimodal mastery value is lower than the first threshold and the correction cost value is higher than the second threshold are marked as high-risk words.

5. The vocabulary comprehensive ability testing method combining spelling and pronunciation according to claim 1, characterized in that: Step S400 includes: S410. Construct a vocabulary association network, wherein the vocabulary association network is a graph structure with words as nodes and word root derivation relations, semantic similarity or phonetic similarity as edges; S420. For each marked high-risk word, retrieve and determine at least one related word that is directly connected to it according to the word association network; S430. Based on the dynamic cognitive state matrix, query and extract the bimodal mastery value and correction cost value of each associated word of the high-risk word, and construct the associated word state set of the high-risk word.

6. The vocabulary comprehensive ability testing method combining spelling and pronunciation according to claim 1, characterized in that: Step S500 includes: S510. For each high-risk word, obtain its associated word status set; S520. Based on the dynamic cognitive state matrix, obtain the baseline bimodal mastery value and baseline correction cost value of the preset word family to which the high-risk words belong; S530. In the set of associated vocabulary states, count the first number of associated vocabulary words whose bimodal mastery value is lower than the benchmark bimodal mastery value, and the second number of associated vocabulary words whose correction cost value is higher than the benchmark correction cost value. S540. Aggregate the first quantity and the second quantity to obtain the number of risk words; if the proportion of the number of risk words to the total number of words in the associated word state set exceeds a preset proportion threshold, it is determined that there is a risk of instability in the word network with the high-risk words as the core.

7. The vocabulary comprehensive ability testing method combining spelling and pronunciation according to claim 1, characterized in that: Step S600 includes: S610. Extract high-risk words and their corresponding bimodal mastery values ​​and correction cost values ​​from the dynamic cognitive state matrix, and extract the list of related words and actual risk ratios of each unstable network from the judgment results of the unstable risk of the word network. S620. Compare the actual risk ratio with multiple preset risk level thresholds to determine the corresponding risk level identifier; S630 integrates high-risk vocabulary, bimodal mastery values, revised cost values, a list of related vocabulary, actual risk ratios, and risk level identifiers to generate a structured diagnostic report.

8. A vocabulary comprehensive ability testing system combining spelling and pronunciation, applied to the vocabulary comprehensive ability testing method combining spelling and pronunciation as described in any one of claims 1-7, characterized in that: The system includes: a data acquisition module, a data analysis module, and a report generation module; The data acquisition module is used to collect the user's spelling data, pronunciation data, and correction operation data for the target words; the correction operation data includes the correction type, the time of correction, and the content of correction. The data analysis module is used to calculate the bimodal mastery value of each target word based on the spelling data and speech data, and to calculate the correction cost value of each target word based on the correction operation data; it is used to construct a dynamic cognitive state matrix based on the bimodal mastery value and the correction cost value, and to mark words with bimodal mastery values ​​below a first threshold and correction cost values ​​above a second threshold as high-risk words; it is used to analyze the state of associated words of the high-risk words based on the word association network to determine the risk of word network instability. The report generation module is used to generate a diagnostic report based on the dynamic cognitive state matrix and the determination result of the risk of instability of the vocabulary network.

9. A vocabulary comprehensive ability testing system combining spelling and pronunciation according to claim 8, characterized in that: The data analysis module includes a state assessment unit and a network analysis unit; The state assessment unit is used to calculate the bimodal mastery value and the correction cost value, construct the dynamic cognitive state matrix, and mark words in the dynamic cognitive state matrix whose bimodal mastery value is lower than the first threshold and whose correction cost value is higher than the second threshold as high-risk words. The network analysis unit is used to analyze the cognitive state of the associated words of the high-risk words based on the word association network, and to determine the risk of instability of the word network based on the deterioration ratio of the state of the associated words.

10. A vocabulary comprehensive ability testing system combining spelling and pronunciation according to claim 8, characterized in that: The report generation module includes a data integration unit and a report synthesis unit; The data integration unit is used to extract high-risk vocabulary information, quantitative indicators, and risk network composition data from the dynamic cognitive state matrix and network instability risk judgment results. The report synthesis unit is used to assess the risk level of the extracted data and generate a structured diagnostic report.

Images