Courseware making method and device based on artificial intelligence and storage medium

Abstract

The invention relates to the technical field of artificial intelligence, in particular to a courseware making method and device based on artificial intelligence and a storage medium, and the method comprises the steps: constructing a tone problem distribution table based on tone accuracy scores and machine bias labels; determining a target training group based on the tone question distribution table, further determining a courseware generation parameter, extracting a recent continuous practice sequence based on the target training group, and further generating a training intensity adjustment parameter to adjust the courseware generation parameter; constructing an individual cognitive bias error model based on listen-and-read entries in a user historical period, and generating a bias error pair list; generating training content based on the courseware generation parameters, the target training group, the bias error pair list and the tone problem distribution table; and arranging a teaching process based on the training content, and generating a courseware script. According to the invention, the pronunciation data, the personal bias mode and the linguistic knowledge base of the learner can be dynamically combined, and an accurate closed loop from data diagnosis to targeted teaching is realized.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence technology, and in particular to a method, apparatus and storage medium for creating courseware based on artificial intelligence. Background Technology

[0002] In online Chinese instruction for native English-speaking adult learners at the beginner level, mastering tone is a major challenge. Traditional online teaching platforms often face the following limitations in tone training: first, the practice materials are mostly fixed and standardized courseware, lacking specificity for individual learners' pronunciation errors; second, they rely on teachers manually listening to and judging errors, which is inefficient and difficult to scale; and third, feedback is delayed, failing to provide immediate and accurate correction during practice.

[0003] While some platforms have introduced automatic voice scoring, most only provide simple scores and fail to deeply diagnose error types and generate progressive training content accordingly.

[0004] Therefore, there is an urgent need for a solution that can automatically analyze learners' pronunciation errors, dynamically generate personalized training courseware, and realize intelligent teaching process management in order to improve the effectiveness and efficiency of tone teaching. Summary of the Invention

[0005] The purpose of this invention is to provide a method, apparatus and storage medium for creating courseware based on artificial intelligence, so as to solve at least one of the problems existing in the prior art.

[0006] To achieve the above objectives, the present invention adopts the following technical solution:

[0007] An artificial intelligence-based courseware production method includes:

[0008] A tone problem distribution table is constructed based on tone accuracy scores and machine error labels;

[0009] Based on the tone problem distribution table, the target training group is determined, and then the courseware generation parameters are determined. Based on the target training group, the recent continuous practice sequence is extracted, and then the training intensity adjustment parameters are generated to adjust the courseware generation parameters.

[0010] Based on the reading entries within the user's historical period, an individual cognitive bias model is constructed, and a list of bias pairs is generated.

[0011] Training content is generated based on courseware generation parameters, target training groups, error pair lists, and tone problem distribution tables.

[0012] The teaching process is designed based on the training content, and courseware scripts are generated.

[0013] Furthermore, the reading entries from the user's historical period are read and grouped according to two dimensions: target tone and syllable position. The target tone includes first tone, second tone, third tone, fourth tone and neutral tone. The syllable position includes single character, first character of two-character words, and last character of two-character words.

[0014] For each group, calculate the average tone accuracy score of all the reading items in that group, and denot it as the average accuracy value Zp;

[0015] The percentage of reading entries with tone accuracy scores below a preset accuracy threshold in this group is recorded as the error rate Cf.

[0016] The machine error labels of all entries in this group whose tone accuracy scores are lower than a preset accuracy threshold are mapped to specific error types, and the error type that occurs most frequently is counted as the main error type of this group.

[0017] The average accuracy value, error rate, and main error types are used as the distribution table for tone problems.

[0018] Furthermore, the error rates of all groups in the tone problem distribution table are sorted from high to low, and each group in the sort is filtered. After filtering, the group with the highest error rate is selected and determined as the target training group for this courseware generation.

[0019] The filtering process is as follows:

[0020] If the average accuracy value of the group is greater than or equal to the preset accuracy filtering threshold or the training status of the group in the teaching schedule is not open, then the group will be filtered; otherwise, the group will not be filtered.

[0021] The courseware generation parameters are calculated based on the average accuracy value A of the target training group. The courseware generation parameters include the number of contrast word pairs N and the exercise format level L.

[0022] Furthermore, extract all the reading entries for the target training group in the historical period, sort these reading entries in ascending order according to their timestamps to form a recent continuous practice sequence, and denot the length of the recent continuous practice sequence as Nt;

[0023] When Nt is greater than the preset length, the continuous practice sequence that meets the conditions is divided into three continuous subsequences in chronological order. The subsequences include an early segment, a middle segment, and a late segment. The early segment consists of the first floor(Nt / 3) follow-up reading items, the middle segment consists of the next floor(Nt / 3) follow-up reading items, and the late segment consists of all remaining items. Here, floor() is a floor function.

[0024] Calculate the average tone accuracy score of all entries in each subsequence, denoted as Me, Mm, and Ml;

[0025] Calculate the average difference Δm between the later segment and the earlier segment, where Δm = Ml - Me;

[0026] If Δm is less than the preset difference, it is determined that there is a fatigue attenuation trend; otherwise, it is determined that there is no fatigue attenuation trend.

[0027] When there is a fatigue decay trend, the training intensity adjustment parameter U is set to [1-min(1,max(0,(-△m) / my))], where my is the preset difference threshold;

[0028] When there is no fatigue decay trend or Nt is less than or equal to the preset length, the training intensity adjustment parameter U is set to 1;

[0029] The number of contrast word pairs N is adjusted based on the training intensity parameter U, and the adjusted number of contrast word pairs N is set to N1, N1=max(2,ceil(N×U)); where ceil() is the floor function.

[0030] Furthermore, the following entries with tone accuracy scores below 80 points in the user's historical period are extracted as sample entries. For each sample entry, the target tone i is paired with the actual tone j according to the error type mapped by its machine error label. All pairings are counted to generate a 5×5 individual tone confusion matrix P. The formula for calculating each element Pij in this matrix is: Pij = number of entries with target tone i that are mispronounced as j / total number of error entries with target tone i.

[0031] The individual tone confusion matrix P is compared element by element with the tone confusion probability matrix M in the linguistic knowledge base. If and only if Pij-Mij>Δk, the item (i,j) is determined to be the user's significant cognitive bias, where Δk is a preset significance threshold.

[0032] Output a list of all salient cognitive biases (i, j) that satisfy the above conditions, denoted as the bias pair list E, which is sorted in descending order of the values ​​of Pij-Mij.

[0033] Furthermore, based on the main error types and error pair list E of the target training group, word pairs that match the tone combinations with items in error pair list E are extracted from the hierarchical minimum tone opposition word library to form a high-priority candidate set; then, based on the number of comparison word pairs N, N sets of word pairs are extracted from the retrieval result list using a random sampling algorithm. If the number is insufficient, it is supplemented from the general matching word library to form the core comparison word pair set of this courseware.

[0034] For each pair of words in the core contrast word pair set, the corresponding template is called to generate a specific practice sentence according to the form specified by the exercise form level L.

[0035] Furthermore, the method for generating the practice sentences includes: if L=1, the practice sentence is the word pair itself;

[0036] If L=2, then each character in the word pair will be embedded into a high-frequency two-character word template to generate a two-character word;

[0037] If L=3, the words in the word pair will be embedded into a preset simple sentence template to generate a simple sentence.

[0038] Furthermore, the training content is arranged into independent basic teaching units according to its order in the word pair set, and a standardized interactive process is designed for each basic teaching unit. After the reading practice section, automated scoring is performed. The automated scoring process is as follows:

[0039] The learner obtains a tone accuracy score for the current practice sentence, denoted as Sc. If Sc is greater than or equal to the first preset score, the learner is deemed to have mastered the current exercise and automatically proceeds to the next basic teaching unit in the sequence. If Sc is greater than or equal to the second preset score but less than the first preset score, the learner is deemed to have partially mastered the current exercise and receives a "Partially correct, please pay attention" message. The process then automatically proceeds to the next basic teaching unit. If Sc is less than the second preset score, the learner is deemed to have difficulty with the current exercise and immediately triggers a preset auxiliary teaching sub-process.

[0040] Integrate the sequences of all the basic teaching units mentioned above and their respective defined automated scoring logic to generate courseware scripts.

[0041] According to another aspect of this application, an artificial intelligence-based courseware creation device is provided, comprising:

[0042] The problem identification unit is used to construct a tone problem distribution table based on tone accuracy scores and machine error labels.

[0043] The parameter determination unit is used to determine the target training group based on the tone problem distribution table, and then determine the courseware generation parameters. It also extracts recent continuous practice sequences based on the target training group to generate training intensity adjustment parameters to adjust the courseware generation parameters.

[0044] The error pair determination unit is used to build an individual cognitive error model based on the reading items within the user's historical period and generate a list of error pairs.

[0045] The training content generation unit is used to generate training content based on courseware generation parameters, target training groups, error pair lists, and tone problem distribution tables.

[0046] The courseware generation unit is used to arrange the teaching process based on the training content and generate courseware scripts.

[0047] According to another aspect of this application, a computer-readable storage medium is provided, the computer-readable storage medium storing a computer program, wherein the computer program is used to control the electronic device on which the computer-readable storage medium is located to execute the aforementioned artificial intelligence-based courseware production method during runtime.

[0048] The beneficial effects of this invention are as follows: The AI-based courseware production method provided in this embodiment brings a systematic innovation to online Chinese tone teaching. Through multi-source data fusion and intelligent analysis, it constructs a complete automated closed loop from problem diagnosis and content generation to process delivery. This method can generate truly personalized dynamic training plans for each learner, accurately focusing on their unique tone confusion patterns and real-time learning status, and automatically adapting practice intensity and teaching strategies. This not only significantly improves the efficiency and relevance of tone training and reduces repetitive work for teachers, but more importantly, it enhances learners' sense of participation and accomplishment through immediate and accurate feedback and personalized content presentation. It provides a replicable and scalable efficient paradigm for automated language skills teaching, powerfully promoting the in-depth development of online education towards intelligence and personalization. Attached Figure Description

[0049] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0050] Figure 1 This is a flowchart illustrating the AI-based courseware creation method of this embodiment.

[0051] Figure 2 This is a flowchart illustrating the method for determining the courseware generation parameters in this embodiment.

[0052] Figure 3 This is a schematic diagram of the structure of the AI-based courseware production device in this embodiment.

[0053] Figure 4 This is a schematic diagram of the electronic device in this embodiment. Detailed Implementation

[0054] To more clearly illustrate the present invention, the following description, in conjunction with preferred embodiments and accompanying drawings, further explains the invention. Similar components in the drawings are indicated by the same reference numerals. Those skilled in the art should understand that the specific description below is illustrative rather than restrictive and should not be construed as limiting the scope of protection of the present invention.

[0055] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate for the embodiments of this application described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0056] Specifically, this embodiment is applied to a beginner-level one-on-one online Chinese teaching platform for adult learners whose native language is English, and is specifically designed to automatically generate training courseware tailored to individual pronunciation tones.

[0057] Please see Figure 1 The diagram shown is a flowchart illustrating the AI-based courseware creation method of this embodiment. Before the method is executed, the system synchronously collects multi-source data through the data interface deployed on the teaching platform, including:

[0058] The learner collects the reading audio through the microphone of the learning terminal and processes it by calling the API interface of the cloud-based acoustic scoring engine to obtain the tone accuracy score and machine error label for each syllable. The tone accuracy score ranges from 0 to 100. The tone accuracy score output by the cloud-based acoustic scoring engine is generated by a standardized mapping of the confidence level output by the time-series classification model. For a target syllable, the model outputs a probability set {p1, p2, p3, p4, p0} of its belonging to five tone categories (first tone, second tone, third tone, fourth tone, and neutral tone). Let the probability corresponding to the target tone be ptarget, then the tone accuracy score Score = 100 × ptarget. This linear mapping directly converts the model's confidence level in judging the target tone into a percentage score. The higher the score, the closer the pronunciation is to the standard. Those skilled in the art can also use other non-linear mappings (such as those based on the Sigmoid function) to achieve this. This specific formula is only a clear and effective preferred implementation scheme.

[0059] The "machine error label" returned by the cloud-based acoustic scoring engine API is an encoded value that represents the most likely tone error identified by the engine. The system internally maintains a mapping table to decode this label into a specific "error type." For example, the label T2_AS_T3 might map to "error type: second tone pronounced as third tone (rising tone pronounced as falling-rising tone)." The cloud-based acoustic scoring engine is based on a deep learning model, such as a recurrent neural network with connectionist temporal classification loss or a Transformer model. Its internal processing flow includes: preprocessing the input audio and extracting the packet... The acoustic feature sequence, including fundamental frequency and Mel-frequency cepstral coefficients, is input into a time-series classification model trained on a large amount of standard Mandarin speech data. The model outputs the tone category (first tone, second tone, third tone, fourth tone, neutral tone) and its confidence score for each syllable. The identified tone is compared with the target tone, and the tone accuracy score can be directly mapped using the confidence score of the recognition model. Machine error labels are generated when the identified tone is inconsistent with the target tone. The encoding rule is T[target]_AS_T[recognition]. For example, if the target second tone is misidentified as the third tone, the label T2_AS_T3 is generated.

[0060] From the learner model database, retrieve the learner's historical performance records of the most recent N effective training sessions. An effective training session refers to a session in which the learner completes all preset exercises and the audio quality passes the signal-to-noise ratio test. The records are stored in a structured format of tuples {training objective, exercise, final score}. In this embodiment, N=50.

[0061] From a linguistic knowledge base, a tone confusion probability matrix for native English speakers and a hierarchical minimum tone opposition word list are retrieved. The tone confusion probability matrix is ​​a 5×5 matrix, where element Mij represents the statistical probability of perceiving or pronouncing the target tone i as tone j. The linguistic knowledge base is a pre-constructed structured database, including at least:

[0062] (1) Tone confusion probability matrix M: obtained by statistical learning of a large number of pronunciation error data of native English speakers learning Chinese. The matrix dimension is 5×5. The row index i represents the target tone and the column index j represents the tone that is mispronounced. The element Mij is calculated as the proportion of all pronunciation error samples with target tone i that are identified as tone j.

[0063] Graded Minimal Tone Contrast Pair Lexicon: This lexicon is stored by grading according to lexical difficulty, such as HSK levels 1-3. Each entry is a set of "minimal contrast pairs", that is, monosyllabic characters or words with exactly the same initials and finals except for different tones. For example, under HSK level 1, {'pair_id': 'T4_T0_001', 'char1': {'character': '子', 'pinyin': 'zǐ', 'tone': 4}, 'char2': {'character': '子', 'pinyin': 'zi', 'tone': 0}} (such as '子' in '儿子')... The construction of the lexicon is based on the commonly used vocabulary list in teaching Chinese as a foreign language and is annotated by linguistics experts;

[0064] (3)General Matching Lexicon: As a supplement to the Minimal Tone Contrast Pair Lexicon, it stores a more extensive syllable-character mapping table containing the target tone and its common word examples. For example, when looking for practice characters for the target tone "third tone", if there are insufficient matches in the Minimal Contrast Pair Lexicon, a common third-tone character, such as "好 hǎo", and one of its common collocations, such as "好看", are randomly selected from this lexicon;

[0065] From the multimedia resource library, index standard pronunciation video clips, static tongue position diagrams, and interactive tone contour canvas components through the Uniform Resource Identifier indexing standard; the multimedia resource library indexes the following prefabricated resources through URI:

[0066] Standard Pronunciation Video Clips: Recorded by professional announcers, each video corresponds to the standard pronunciation of a specific Chinese character or word, including a close-up of the front mouth shape and a display of the side tongue position;

[0067] Static Tongue Position Diagrams: Provide vector diagrams of the positions of the speech organs for each combination of initials, finals, and tones;

[0068] Interactive Tone Contour Canvas Component: A Web or mobile UI component that can receive fundamental tone frequency (F0) data and draw a five-degree marking diagram (such as a pitch trajectory from 1 to 5) for visualizing the comparison between the standard tone curve and the user's pronunciation curve.

[0069] In this embodiment, no specific limitations are imposed on the acquisition, storage, and API call methods of the above data, and those skilled in the art can freely set them according to requirements.

[0070] Specifically, this embodiment provides a cold start teaching plan for new users or learners with insufficient data: When the number of valid follow-up records is less than 10, the system preferentially trains the pronunciation of single characters in the first and fourth tones, uses 3 groups of contrast word pairs for single character practice, and as the data accumulates, the system gradually transitions to a fully personalized mode after 5-15 training sessions.

[0071] The method includes:

[0072] Step S1: Construct a tone problem distribution table based on tone accuracy scores and machine error labels.

[0073] Specifically, the system reads the reading entries from the user's historical period and groups them according to two dimensions: target tone and syllable position. The target tone includes the first tone, second tone, third tone, fourth tone, and neutral tone. The syllable position includes single characters, the first character of two-character words, and the last character of two-character words. The reading entry is the smallest learning unit recorded by the system, which refers to a complete reading of a single target pronunciation by the user. It includes data such as user audio, tone accuracy score, and error type.

[0074] For each group, calculate the average tone accuracy score of all the reading items in that group, and denot it as the average accuracy value Zp;

[0075] The percentage of reading entries with tone accuracy scores below a preset accuracy threshold in this group is recorded as the error rate Cf.

[0076] The machine error labels of all entries in this group whose tone accuracy scores are lower than a preset accuracy threshold are mapped to specific error types, and the error type that occurs most frequently is counted as the main error type of this group.

[0077] The average accuracy value, error rate, and main error types are used as the distribution table for tone problems.

[0078] Specifically, in this embodiment, the preset accuracy threshold is 75 points.

[0079] Specifically, by systematically analyzing learners' historical pronunciation data, abstract pronunciation problems are transformed into structured distribution charts. This helps the system go beyond a single score and accurately pinpoint learners' weaknesses from multiple dimensions, such as target tone and syllable position.

[0080] Please continue reading. Figure 1 As shown, the AI-based courseware production method further includes:

[0081] Step S2: Determine the target training group based on the tone problem distribution table, then determine the courseware generation parameters, and extract recent continuous practice sequences based on the target training group to generate training intensity adjustment parameters to adjust the courseware generation parameters.

[0082] Specifically, based on the learner's specific problem distribution, the system intelligently determines the training focus, the complexity and quantity of exercises, ensuring that the difficulty of the courseware matches the learner's current level. On the other hand, by analyzing the performance trend of recent practice sequences, the system can keenly identify the "fatigue decay" phenomenon that learners may experience, and automatically adjust the training intensity accordingly to help maintain learning motivation and achieve a scientific training rhythm that is both relaxed and focused.

[0083] Please see Figure 2 As shown, the method for determining the courseware generation parameters includes:

[0084] Step S21: Determine the target training group based on the tone problem distribution table, and then determine the courseware generation parameters.

[0085] Specifically, the error rates of all groups in the tone problem distribution table are sorted from high to low, and each group in the sort is filtered. After filtering, the group with the highest error rate is selected and determined as the target training group for this courseware generation.

[0086] The filtering process is as follows:

[0087] If the average accuracy value of the group is greater than or equal to the preset accuracy filtering threshold or the training status of the group in the teaching schedule is not open, then the group will be filtered; otherwise, the group will not be filtered.

[0088] The courseware generation parameters are calculated based on the average accuracy value A of the target training group. The courseware generation parameters include the number of contrast word pairs N and the exercise format level L.

[0089] The formula for calculating the number N of the contrasting word pairs is:

[0090] N=max(2,min(5,int((100-A) / 15)))

[0091] Here, int() is the integer function;

[0092] The formula for calculating the exercise level L is:

[0093] First, determine the initial level based on the syllable position in the target training group: if it is a single character, then Lc=1; if it is the first character or the last character of a two-character word, then Lc=2.

[0094] Secondly, the initial level is adjusted based on the error rate of the target training group to determine the practice format level L: if the error rate of the group is greater than 0.7, then the practice format level L = max(1, Lc-1); if the average accuracy value of the group is greater than the preset high mastery threshold and the error rate Cf is lower than the preset low error rate threshold, then the practice format level L = min(3, Lc+1); if the above conditions are not met, then the practice format level L = Lc.

[0095] Where L=1 indicates that the practice material is a single character; L=2 indicates that the practice material is a two-character word; and L=3 indicates that the practice material is a short sentence.

[0096] Specifically, in this embodiment, the preset accuracy filtering threshold is 85 points, the preset high mastery threshold is 90 points, and the preset low error rate threshold is 0.15.

[0097] Specifically, the teaching schedule is a structured data record of the course syllabus, defining the overall sequence of tone teaching and the "open status" of each tone-syllable position combination. For example, {"tone":1, "position":"single", "status":"completed"} indicates that the teaching of a single word with the first tone has been completed; {"tone":3, "position":"word_end", "status":"locked"} indicates that the teaching unit for the third tone at the end of a two-character word has not yet been unlocked. This table is pre-set by the teacher or system administrator according to the teaching plan.

[0098] Please continue reading. Figure 2 As shown, the method for determining the courseware generation parameters further includes:

[0099] Step S22: Extract recent continuous practice sequences based on the target training group, and then generate training intensity adjustment parameters to adjust the courseware generation parameters.

[0100] Specifically, extract all the reading entries for the target training group in the historical period, sort these reading entries in ascending order according to their timestamps to form a recent continuous practice sequence, and denot the length of the recent continuous practice sequence as Nt;

[0101] When Nt is greater than the preset length, the continuous practice sequence that meets the conditions is divided into three continuous subsequences in chronological order. The subsequences include an early segment, a middle segment, and a late segment. The early segment consists of the first floor(Nt / 3) follow-up reading items, the middle segment consists of the next floor(Nt / 3) follow-up reading items, and the late segment consists of all remaining items. Here, floor() is a floor function.

[0102] Calculate the average tone accuracy score of all entries in each subsequence, denoted as Me, Mm, and Ml;

[0103] Calculate the average difference Δm between the later segment and the earlier segment, where Δm = Ml - Me;

[0104] If Δm is less than the preset difference, it is determined that there is a fatigue attenuation trend; otherwise, it is determined that there is no fatigue attenuation trend.

[0105] When there is a fatigue decay trend, the training intensity adjustment parameter U is set to [1-min(1,max(0,(-△m) / my))], where my is the preset difference threshold;

[0106] When there is no fatigue decay trend or Nt is less than or equal to the preset length, the training intensity adjustment parameter U is set to 1;

[0107] The number of contrast word pairs N is adjusted based on the training intensity parameter U, and the adjusted number of contrast word pairs N is set to N1, N1=max(2,ceil(N×U)); where ceil() is the floor function.

[0108] Specifically, in this embodiment, the preset length is 15, the preset difference is -5, the preset difference threshold is 30, and the historical period is 15 days.

[0109] Please continue reading. Figure 1 As shown, the AI-based courseware production method further includes:

[0110] Step S3: Construct an individual cognitive bias model based on the reading items within the user's historical period and generate a list of bias pairs.

[0111] Specifically, the following entries with tone accuracy scores below 80 points in the user's historical period are extracted as sample entries. For each sample entry, the target tone i is paired with the actual tone j according to the error type mapped by its machine error label. All pairings are counted to generate a 5×5 individual tone confusion matrix P. The formula for calculating each element Pij in this matrix is: Pij = number of entries with target tone i that are mispronounced as j / total number of error entries with target tone i.

[0112] The individual tone confusion matrix P is compared element by element with the tone confusion probability matrix M in the linguistic knowledge base. If and only if Pij-Mij>Δk, the item (i,j) is determined to be the user's significant cognitive bias, where Δk is a preset significance threshold.

[0113] Output a list of all salient cognitive biases (i, j) that satisfy the above conditions, denoted as the bias pair list E, which is sorted in descending order of the values ​​of Pij-Mij.

[0114] Specifically, in this embodiment, the preset significance threshold is 0.15.

[0115] Specifically, by comparing learners' individual error patterns with the general confusion patterns of the group, it is possible to accurately identify the learner's particularly significant and personalized tone confusion tendency compared to other native English speakers. This generates an "error pair list," allowing teaching interventions to focus on the learner's unique cognitive difficulties. This enables the allocation of more valuable practice time to address the learner's most deeply ingrained error habits, greatly improving the accuracy and efficiency of personalized teaching.

[0116] Please continue reading. Figure 1 As shown, the AI-based courseware production method further includes:

[0117] Step S4: Generate training content based on courseware generation parameters, target training groups, error pair list and tone problem distribution table.

[0118] Specifically, based on the main error types and error pair list E of the target training group, word pairs that match the tone combinations with items in error pair list E are extracted from the hierarchical minimum tone opposition word library to form a high-priority candidate set; then, based on the number of comparison word pairs N, N sets of word pairs are extracted from the retrieval result list using a random sampling algorithm. If the number is insufficient, it is supplemented from the general matching word library to form the core comparison word pair set of this courseware.

[0119] For each word pair in the core contrast word pair set, based on the format specified by exercise format level L, the corresponding template is invoked to generate a specific exercise sentence:

[0120] If L=1, the exercise sentence is the word pair itself;

[0121] If L=2, then each character in the word pair will be embedded into a high-frequency two-character word template to generate a two-character word;

[0122] If L=3, the words in the word pair will be embedded into a preset simple sentence template to generate a simple sentence.

[0123] Specifically, by utilizing a structured linguistic knowledge base, the system intelligently matches the most relevant contrastive word pairs for the target tone and specific error type, and embeds them into appropriate contexts according to the predetermined practice level. This ensures that the generated training content not only closely matches the learner's current problems, but also conforms to their language proficiency stage, making the automatically generated courseware both targeted and pedagogically reasonable.

[0124] Specifically, the system maintains a teaching template library to generate practice sentences based on the exercise format level L:

[0125] Two-character word templates: These are a set of high-frequency, fixed-structure two-character word frames, with one empty space reserved for embedding the target character. For example, the template "very [X]" can be used to embed the target character "busy (máng)" to generate "very busy". The template library is classified according to the part of speech of the target character (such as adjective or verb).

[0126] Simple sentence templates: These are a set of basic Chinese sentence patterns with key blanks, for example, the template "I like to eat [X]" or "This is [X]". These templates are selected from the HSK Elementary Grammar Outline.

[0127] Please continue reading. Figure 1 As shown, the AI-based courseware production method further includes:

[0128] Step S5: Arrange the teaching process based on the training content and generate courseware scripts.

[0129] Specifically, the training content is arranged into independent basic teaching units according to its order in the word pair set. A standardized interactive process is designed for each basic teaching unit, and automated scoring is performed after its follow-up reading practice. The automated scoring process is as follows:

[0130] The learner obtains a tone accuracy score (Sc) for the current practice sentence. If Sc is greater than or equal to a first preset score, the learner is deemed to have mastered the current exercise and automatically proceeds to the next basic teaching unit in the sequence. If Sc is greater than or equal to a second preset score but less than the first preset score, the learner is deemed to have partially mastered the current exercise and receives a "Partially correct, please pay attention" message. The process then automatically proceeds to the next basic teaching unit. If Sc is less than the second preset score, the learner is deemed to have difficulty with the current exercise and immediately triggers a preset auxiliary teaching sub-process. The preset auxiliary teaching sub-process includes, but is not limited to, displaying the error correction prompt text associated with the current unit in a prominent position on the interface and automatically playing the standard pronunciation video of the current unit at least once.

[0131] Integrate the sequence of all the basic teaching units mentioned above and their respective defined automated scoring logic to generate a courseware script. The script is a machine-parseable data structure, such as JSON or XML format, which clearly defines all the teaching content of the courseware, the fixed order of exercises, the judgment conditions for each exercise, and the specific teaching behaviors that should be triggered under different conditions.

[0132] Specifically, personalized training content is encapsulated into a standardized, automatically executed teaching interaction process. It incorporates an automated decision-making logic based on real-time scoring, which adjusts the teaching path in real time according to learners' performance: progression is advanced if mastered, prompts are provided for partial mastery, and auxiliary teaching sub-processes are immediately triggered for difficulties, ensuring the continuity and interactivity of the teaching process. The resulting standardized courseware scripts enable the teaching platform to deliver and execute highly personalized teaching content stably and consistently, achieving a unity of large-scale automation and deep personalization.

[0133] Specifically, in this embodiment, the first preset score is 85 points and the second preset score is 65 points.

[0134] For example, the specific logic of the auxiliary teaching sub-process can be as follows: When it is determined that the current exercise is difficult, the auxiliary teaching sub-process is triggered as follows:

[0135] The system selects and combines a prompt from a predefined prompt library based on the target tone of the current exercise and the most relevant item in the user's "main error type" or "error pair list" in this exercise unit. For example, for the error of pronouncing the target tone "second tone" as "first tone", the prompt might be: "Please note that the second tone is a rising tone. Please raise the tone from the middle, instead of keeping it level."

[0136] In the pop-up interface, three items are displayed simultaneously: (i) the above error correction prompt text; (ii) the standard pronunciation video of the current Chinese character is played automatically (at least once); (iii) on the interactive tone canvas, the rising curve of the standard second tone and the flat curve of the current user's pronunciation are drawn side by side for visual comparison.

[0137] After completing the above auxiliary demonstration, the system pauses and waits for the user to click the "Try Again" button. After the user clicks, the system re-records and scores. If the new score is still lower than the second preset score, the auxiliary process can be repeated or the difficulty can be marked for subsequent manual intervention by the teacher. The maximum number of auxiliary triggers is 3.

[0138] Please see Figure 3 As shown, the AI-based courseware creation device includes:

[0139] The problem identification unit is used to construct a tone problem distribution table based on tone accuracy scores and machine error labels.

[0140] The parameter determination unit is used to determine the target training group based on the tone problem distribution table, and then determine the courseware generation parameters. It also extracts recent continuous practice sequences based on the target training group to generate training intensity adjustment parameters to adjust the courseware generation parameters.

[0141] The error pair determination unit is used to build an individual cognitive error model based on the reading items within the user's historical period and generate a list of error pairs.

[0142] The training content generation unit is used to generate training content based on courseware generation parameters, target training groups, error pair lists, and tone problem distribution tables.

[0143] The courseware generation unit is used to arrange the teaching process based on the training content and generate courseware scripts.

[0144] The AI-based courseware production device provided in this application embodiment can execute the AI-based courseware production method provided in any embodiment of this application, and has the corresponding functional modules and beneficial effects of the execution method.

[0145] Please see Figure 4 As shown, it is a structural schematic diagram of an electronic device in this embodiment. The electronic device in this embodiment may include, but is not limited to, mobile terminals such as mobile phones, laptops, digital broadcast receivers, PDAs (personal digital assistants), PADs (tablet computers), PMPs (portable multimedia players), vehicle terminals (such as vehicle navigation terminals), wearable electronic devices, etc., as well as fixed terminals such as digital TVs, desktop computers, smart home devices, etc. Figure 4 The electronic device shown is merely an example and should not be construed as limiting the functionality and scope of use of the embodiments of the present invention.

[0146] like Figure 4 As shown, the electronic device includes: a processor 501, a memory 502, a communication interface 503, and a system bus 504. The processor includes at least one of a central processing unit (CPU), a graphics processing unit (GPU), or a field-programmable gate array (FPGA), configured to call computer programs and data stored in the memory and generate control instructions; the memory includes random access memory (RAM) and / or non-volatile memory (NVM), the NVM including flash memory, solid-state drive (SSD), or a combination thereof, used to store computer programs, process intermediate data, and historical data sets; the communication interface includes a wired communication module and a wireless communication module, the wired communication module supporting Ethernet or RS-485 protocols for connecting to sensor networks; the wireless communication module supporting LoRa, 5G, or satellite communication protocols for transmitting processing results to a remote server; the system bus adopts a PCI Express or AXI bus architecture to achieve high-speed data interaction and clock synchronization between the processor, memory, and communication interface.

[0147] This embodiment also provides a computer-readable storage medium, which physically stores computer-executable instructions. When the instructions are transmitted to the processing unit via the integrated circuit substrate, they are encapsulated and processed through the data channel of the bus system and then solidified into the non-volatile storage area of ​​the storage module. The executable instructions are configured to implement the complete technical solution of the artificial intelligence-based courseware production method when executed by the processor.

[0148] Obviously, the above embodiments of the present invention are merely examples for clearly illustrating the present invention, and are not intended to limit the implementation of the present invention. For those skilled in the art, other variations or modifications can be made based on the above description. It is impossible to exhaustively list all the implementation methods here. All obvious variations or modifications derived from the technical solutions of the present invention are still within the protection scope of the present invention.

Claims

1. A method for creating courseware based on artificial intelligence, characterized in that, include: A tone problem distribution table is constructed based on tone accuracy scores and machine error labels; Based on the tone problem distribution table, the target training group is determined, and then the courseware generation parameters are determined. Based on the target training group, the recent continuous practice sequence is extracted, and then the training intensity adjustment parameters are generated to adjust the courseware generation parameters. Based on the reading entries within the user's historical period, an individual cognitive bias model is constructed, and a list of bias pairs is generated. Training content is generated based on courseware generation parameters, target training groups, error pair lists, and tone problem distribution tables. The teaching process is designed based on the training content, and courseware scripts are generated.

2. The method for creating courseware based on artificial intelligence according to claim 1, characterized in that, Read the reading entries from the user's historical period and group them according to two dimensions: target tone and syllable position. The target tone includes first tone, second tone, third tone, fourth tone and neutral tone. The syllable position includes single character, first character of two-character words, and last character of two-character words. For each group, calculate the average tone accuracy score of all the reading items in that group, and denot it as the average accuracy value Zp; The percentage of reading entries with tone accuracy scores below a preset accuracy threshold in this group is recorded as the error rate Cf. The machine error labels of all entries in this group whose tone accuracy scores are lower than a preset accuracy threshold are mapped to specific error types, and the error type that occurs most frequently is counted as the main error type of this group. The average accuracy value, error rate, and main error types are used as the distribution table for tone problems.

3. The method for creating courseware based on artificial intelligence according to claim 2, characterized in that, Sort all groups in the tone problem distribution table by error rate from high to low, filter each group in the sort, and select the group with the highest error rate as the target training group for this courseware generation. The filtering process is as follows: If the average accuracy value of the group is greater than or equal to the preset accuracy filtering threshold or the training status of the group in the teaching schedule is not open, then the group will be filtered; otherwise, the group will not be filtered. The courseware generation parameters are calculated based on the average accuracy value A of the target training group. The courseware generation parameters include the number of contrast word pairs N and the exercise format level L.

4. The method for creating courseware based on artificial intelligence according to claim 3, characterized in that, Extract all the reading entries for the target training group in the historical period, sort these reading entries in ascending order according to their timestamps to form a recent continuous practice sequence, and denote the length of the recent continuous practice sequence as Nt; When Nt is greater than the preset length, the continuous practice sequence that meets the conditions is divided into three continuous subsequences in chronological order. The subsequences include an early segment, a middle segment, and a late segment. The early segment consists of the first floor(Nt / 3) follow-up reading items, the middle segment consists of the next floor(Nt / 3) follow-up reading items, and the late segment consists of all remaining items. Here, floor() is a floor function. Calculate the average tone accuracy score of all entries in each subsequence, denoted as Me, Mm, and Ml; Calculate the average difference Δm between the later segment and the earlier segment, where Δm = Ml - Me; If Δm is less than the preset difference, it is determined that there is a fatigue attenuation trend; otherwise, it is determined that there is no fatigue attenuation trend. When there is a fatigue decay trend, the training intensity adjustment parameter U is set to [1-min(1,max(0,(-△m) / my))], where my is the preset difference threshold; When there is no fatigue decay trend or Nt is less than or equal to the preset length, the training intensity adjustment parameter U is set to 1; The number of contrast word pairs N is adjusted based on the training intensity parameter U, and the adjusted number of contrast word pairs N is set to N1, N1=max(2,ceil(N×U)); where ceil() is the floor function.

5. The method for creating courseware based on artificial intelligence according to claim 4, characterized in that, Extract reading entries with tone accuracy scores below 80 from the user's historical reading period as sample entries. For each sample entry, pair the target tone i with the actual tone j according to the error type mapped by its machine error label. Count all pairings and generate a 5×5 individual tone confusion matrix P. The formula for calculating each element Pij in this matrix is: Pij = number of entries with target tone i that are mispronounced as j / total number of error entries with target tone i. The individual tone confusion matrix P is compared element by element with the tone confusion probability matrix M in the linguistic knowledge base. If and only if Pij-Mij>Δk, the item (i,j) is determined to be the user's significant cognitive bias, where Δk is a preset significance threshold. Output a list of all salient cognitive biases (i, j) that satisfy the above conditions, denoted as the bias pair list E, which is sorted in descending order of the values ​​of Pij-Mij.

6. The method for creating courseware based on artificial intelligence according to claim 5, characterized in that, Based on the main error types and error pair list E of the target training group, word pairs that match the tone combinations with items in error pair list E are extracted from the hierarchical minimum tone opposition word library to form a high-priority candidate set; then, based on the number of comparison word pairs N, N sets of word pairs are extracted from the retrieval result list using a random sampling algorithm. If the number is insufficient, it is supplemented from the general matching word library to form the core comparison word pair set of this courseware. For each pair of words in the core contrast word pair set, the corresponding template is called to generate a specific practice sentence according to the form specified by the exercise form level L.

7. The method for creating courseware based on artificial intelligence according to claim 6, characterized in that, The method for generating the practice sentences includes: if L=1, the practice sentence is the word pair itself; If L=2, then each character in the word pair will be embedded into a high-frequency two-character word template to generate a two-character word; If L=3, the words in the word pair will be embedded into a preset simple sentence template to generate a simple sentence.

8. The method for creating courseware based on artificial intelligence according to claim 7, characterized in that, The training content is arranged into independent basic teaching units according to its order in the word pair set. A standardized interactive process is designed for each basic teaching unit, and automated scoring is performed after its reading practice section. The automated scoring process is as follows: The learner obtains a score for the accuracy of tone reading of the current exercise sentence, denoted as Sc. If Sc is greater than or equal to the first preset score, it is determined that the current exercise has been mastered and the process automatically proceeds to the next basic teaching unit in the sequence. If Sc is greater than or equal to the second preset score and less than the first preset score, it is determined that the current exercise is partially mastered and a prompt message of "Partially correct, please pay attention" is provided. The process then automatically proceeds to the next basic teaching unit. If Sc is less than the second preset score, it is determined that the current exercise is difficult, and the preset auxiliary teaching sub-process is immediately triggered; Integrate the sequences of all the basic teaching units mentioned above and their respective defined automated scoring logic to generate courseware scripts.

9. An artificial intelligence-based courseware production device, applied to the artificial intelligence-based courseware production method as described in any one of claims 1-8, characterized in that, include: The problem identification unit is used to construct a tone problem distribution table based on tone accuracy scores and machine error labels. The parameter determination unit is used to determine the target training group based on the tone problem distribution table, and then determine the courseware generation parameters. It also extracts recent continuous practice sequences based on the target training group to generate training intensity adjustment parameters to adjust the courseware generation parameters. The error pair determination unit is used to build an individual cognitive error model based on the reading items within the user's historical period and generate a list of error pairs. The training content generation unit is used to generate training content based on courseware generation parameters, target training groups, error pair lists, and tone problem distribution tables. The courseware generation unit is used to arrange the teaching process based on the training content and generate courseware scripts.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, wherein the computer program is used to control the electronic device on which the computer-readable storage medium is located to perform the artificial intelligence-based courseware production method according to any one of claims 1-8 during runtime.

Images