A concise and simple word intelligent management system and method based on deep learning
Through deep learning technology, based on user vocabulary testing and feature analysis, a personalized word learning path is generated, which solves the problem of neglecting the difference between word count and difficulty in the existing methods, and improves learning efficiency and motivation.
Patent Information
- Application Number
- CN202411008490.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-26
- Publication Date
- 2024-11-15
- Estimated Expiration
- 2044-07-26
AI Technical Summary
Existing word learning methods cannot effectively reduce the number of words that learners need to face, ignore the differences between individual differences and word difficulty, resulting in limited learning efficiency and motivation.
Using a smart and simple word management system based on deep learning, through user vocabulary testing, feature vector analysis and self-attention mechanism model, a personalized list of words to be learned is generated, the learning path is dynamically updated, and learning is adjusted based on user learning records and feedback. content.
A more personalized learning plan has been realized, which significantly reduces learning costs, improves learning efficiency and motivation, shortens the learning path, and makes the learning process more efficient.
Smart Images

Figure CN118535721B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of data management, and in particular to a concise simple word intelligent management system and method based on deep learning. Background Art
[0002] In modern educational technology, personalized learning programs have gradually become a key way to improve learning efficiency, especially in the field of language learning. Traditional word learning methods mainly rely on the Ebbinghaus forgetting curve. Although the learning strategies guided by this theory have rationally arranged the learning content in order to consolidate memory through reasonable repeated review, they mainly focus on how to more effectively plan and sequence word learning, rather than reducing the number of words to be learned.
[0003] Therefore, the existing technology has the following limitations:
[0004] Content sorting rather than quantity reduction: Although existing learning solutions can optimize the time points of learning and reviewing according to the Ebbinghaus forgetting curve, these methods are mainly used to reasonably sort the learning content rather than reduce the total amount of learning. For learners, the amount of learning materials they need to face has not decreased, which directly affects learning efficiency and motivation.
[0005] Ignoring individual differences and word difficulty differences: Traditional methods often use a unified review cycle, ignoring the differences between individual users and the difficulty of words. Different learners have different levels of mastery and learning speed for words of different difficulty levels, and this learning method cannot meet the needs of all users. Summary of the invention
[0006] The purpose of the present invention is to provide a concise simple word intelligent management system and method based on deep learning to solve the problems raised in the above background technology.
[0007] In order to solve the above technical problems, the present invention provides the following technical solutions:
[0008] A concise and simple word intelligent management method based on deep learning, the method comprises the following steps:
[0009] Step S100. The user starts the word learning application through the client for the first time, and the server creates a vocabulary test and generates several test questions; the generated vocabulary test questions are pushed to the client, and the user takes the vocabulary test;
[0010] Step S200. The user completes the vocabulary test, and the client uploads the user's answer data to the server, which analyzes the user's answer data to obtain the user's vocabulary prediction result;
[0011] Step S300. Based on the analysis of the user's answer data by the server, a user vocabulary feature vector is constructed; based on the user vocabulary feature vector, the server obtains the probability distribution of the words that the user has mastered through a deep learning model based on a self-attention mechanism;
[0012] Step S400. The server generates a list of words to be learned based on the predicted results of the probability distribution of the words that the user has mastered; every other day, the list of words to be learned is updated based on the user's learning record of the previous day, thereby obtaining a personalized learning path.
[0013] Step S100 includes:
[0014] S101. The user starts the word learning application through the client for the first time, and the server creates a vocabulary test and generates several test questions; the process of creating a vocabulary test is as follows:
[0015] Get a word database and calculate the difficulty score of each word in the word database. The specific calculation formula for the difficulty score of a word is:
[0016] D=w1*L+w2*(1 / m)*(Σ i∈[1,m] F i )+w3*B,
[0017] Where D represents the difficulty score of the word; w1, w2 and w3 are all weight factors, and w1+W2+w3=1; L represents the length of the word; Fi represents the frequency of the word in the i-th test corpus; m represents the number of tests; B represents the difficulty level of the test with the most frequent word; according to the difficulty score of the word, the words are divided into several vocabulary clusters, and several words are randomly selected from each vocabulary cluster, and test questions are generated based on the selected words;
[0018] S102. The generated test questions are sent to the client in the form of a data packet. The client displays the test questions to the user in order, and the user selects the correct Chinese interpretation for each word displayed; the server records the user's selection in the test in real time. When the vocabulary test reaches a preset number, the user chooses to continue or complete the test.
[0019] Step S200 includes:
[0020] After the user completes the vocabulary test, the client uploads the user's answer data to the server; the server calculates the user data based on word cluster statistics to obtain the user's mastery of words in different word clusters; based on the user's mastery of words in different word clusters, the user's vocabulary S is predicted. The specific calculation formula for the user's vocabulary S is:
[0021] S=Σ c∈Clusters (int(Rc*Vc)),
[0022] Rc=[max(0.3*Kc-Uc)] / (3*Tc),
[0023] Among them, Kc represents the number of known words in word cluster c for the user, Uc represents the number of unknown words in word cluster c for the user, Tc represents the total number of words in word cluster c, and Tc=Kc+Uc+Cc, Cc represents the number of uncertain words in word cluster c for the user; Rc represents the mastery rate of word cluster c, Vc represents the length of the word list in word cluster c; Clusters represents the different word clusters involved in the user's vocabulary test, and each word cluster contains a group of related words.
[0024] The user's vocabulary mastery is assessed through an initial vocabulary test conducted on the client, which involves steps such as creating and pushing the test to the client, recording user selections in real time, and automatically terminating the test based on user responses. The core of this is the ability to dynamically adjust the test length and difficulty based on the user's actual performance until the predetermined confidence rate standard is reached, accurately assessing the user's vocabulary level.
[0025] Step S300 includes:
[0026] S301. For each user, based on the user's answer data, obtain the user's mastery ratio in each word cluster, and construct an n-dimensional feature vector, where n is equal to the number of word clusters; associate each user's n-dimensional feature vector with the corresponding user ID, thereby forming the user vocabulary feature vector V ID , and V ID ={v1,v2,...,vn}, where v1 represents the user's accuracy rate for the first word cluster, v2 represents the user's accuracy rate for the second word cluster, and so on, and vn represents the user's accuracy rate for the nth word cluster;
[0027] Based on the analysis of user test results, their vocabulary level is predicted and this information is converted into vocabulary feature vectors. This involves not only data analysis methods, but also how to effectively store these feature vectors in the server for subsequent customization of learning content. This is a key point in the technical solution because it is directly related to the accuracy and effectiveness of the subsequent personalized learning path.
[0028] S302. The user vocabulary feature vector V ID Perform weighted summation to obtain context coding K. The specific calculation formula of context coding K is:
[0029] K=Σ j∈[1,n] α j V j ,
[0030] Among them, α j represents the weight of the j-th word cluster, V j represents the accuracy of the jth word cluster, and j is a positive integer from 1 to n; j The specific calculation formula is:
[0031] α j =exp(e j ) / [Σ j∈[1,n] exp(e j )],
[0032] e j =ReLU(WqV j ) T (WxV ID ),
[0033] Among them, e j represents the activation value of the jth word cluster; ReLU() represents the rectified linear unit function; Wq and Wx represent parameter matrices, which are used to obtain the attention score according to the feature vector of each word cluster in the user vocabulary feature vector; the context encoding K is used to obtain the prediction result Y of the probability distribution of the words that the user has mastered, and Y=softmax(W0K), where W0 represents the parameter matrix of the output layer; in the above formula, ReLU and softmax are used for the activation function of the neural network and the probability distribution prediction of the output layer, respectively.
[0034] Step S400 includes:
[0035] S401. For each user, the server generates a personalized list of words to be learned based on the predicted results of the probability distribution of the words that the user has mastered; and the generation process of the list of words to be learned is as follows:
[0036] According to the prediction result of the probability distribution of the words that the user has mastered, each word in each word cluster is arranged in descending order according to the probability of the word that the user has mastered, and the words with probability equal to 1 are screened out, and the words with probability equal to 1 are marked as mastered words, and the words with probability less than 0.5 are marked as unmastered words, and the remaining words are marked as words to be consolidated; the marked words are allocated according to the user's learning time and plan, so as to generate a list of words to be learned;
[0037] S402. Every other day, the learning record of the user on the previous day is obtained, and a new user vocabulary feature vector is obtained based on the user's learning record on the previous day. The server passes the new user vocabulary feature vector through a deep learning model based on a self-attention mechanism, thereby obtaining a prediction result of the probability distribution of the new words that the user has mastered, and updates the list of words to be learned based on the prediction result of the probability distribution of the new words that the user has mastered, thereby obtaining a personalized learning path; and the user can delete the latest list of words to be learned output by the client, thereby reducing the time wasted on repeatedly learning known vocabulary and allowing the user to focus on learning new vocabulary.
[0038] During daily learning, the system uses the user's feedback on their mastery of new words to update the user's vocabulary level and adjust subsequent learning content accordingly. This process requires not only real-time data collection and processing, but also the ability to accurately analyze data to ensure that the learning content can accurately match the user's actual needs.
[0039] An intelligent management system for concise simple words based on deep learning, the system includes: a word database management module, a vocabulary test module, a user answer data analysis module, a to-be-learned word list generation and update module, and a learning record management and feedback module;
[0040] The word database management module is responsible for managing the word database, including operations such as word storage, update, and deletion; calculating the difficulty score of each word, and dividing the words into different word clusters according to the score, to provide support for subsequent vocabulary testing and generation of lists of words to be learned;
[0041] The vocabulary test module creates vocabulary test questions based on word clusters, pushes the vocabulary test questions to the client, records the user's answer data, and determines whether to continue or complete the test based on the user's selection and answer situation;
[0042] The user answer data analysis module analyzes the answer data uploaded by the user after completing the vocabulary test to obtain the user's vocabulary prediction result; based on the user's answer data, the user's mastery ratio of each word cluster is obtained to construct the user's vocabulary feature vector;
[0043] The module for generating and updating the list of words to be learned obtains the prediction result of the probability distribution of the words that the user has mastered based on the user vocabulary feature vector, generates the list of words to be learned according to the prediction result of the probability distribution of the words that the user has mastered, and updates the list of words to be learned every day according to the user's learning record;
[0044] The learning record management and feedback module records the user's learning records, including daily learning status, learning plan completion status, etc., and provides feedback and personalized learning path suggestions based on the user's learning status; it also supports users to delete the learning word list, thereby generating a personalized learning path.
[0045] Further, the word database management module includes a database connection unit, a word difficulty score calculation unit and a word cluster generation unit;
[0046] The database connection unit is responsible for establishing a connection with the word database and performing data reading and writing operations; the word difficulty score calculation unit calculates the difficulty score of each word and stores the score result in the database; the word cluster generation unit divides the word into several vocabulary clusters according to the word difficulty score.
[0047] Furthermore, the vocabulary test module includes a test question generating unit and a user answer recording unit;
[0048] The test question generating unit randomly selects words from each vocabulary cluster to generate vocabulary test questions, and pushes the vocabulary test questions to the client according to the user's request when the user first starts the application; the user answer recording unit records the user's choices in the test in real time, and monitors the user's choices after the vocabulary test reaches a preset number.
[0049] Further, the user answer data analysis module includes an answer data uploading unit, a user vocabulary prediction unit, and a user vocabulary feature vector construction unit;
[0050] The answer data uploading unit receives the user answer data uploaded by the client and stores it on the server for analysis; the user vocabulary prediction unit analyzes the user's answer data and predicts the user's vocabulary; the user vocabulary feature vector construction unit obtains the user's mastery ratio of each word cluster based on the user's answer data and constructs a user vocabulary feature vector.
[0051] Further, the to-be-learned word list generation and update module includes a to-be-learned word list generation unit and a to-be-learned word list update unit;
[0052] The to-be-learned word list generating unit obtains the prediction result of the probability distribution of the words that the user has mastered based on the user vocabulary feature vector, and generates the to-be-learned word list according to the prediction result of the probability distribution of the words that the user has mastered; the to-be-learned word list updating unit updates the to-be-learned word list according to the user's learning record of the previous day;
[0053] The learning record management and feedback module includes a learning record management unit and a personalized learning path generation unit;
[0054] The learning record management unit manages and records the user's learning record; the personalized learning path generation unit outputs the updated list of words to be learned to the user, and the user deletes the list of words to be learned to obtain a personalized learning path.
[0055] Compared with the prior art, the beneficial effects achieved by the present invention are:
[0056] The present invention realizes a more personalized learning plan by comprehensively considering the inherent difficulty of words and the individual differences of users. The traditional method only marks words based on whether the user has been exposed to them, while the present invention analyzes in more depth which words are "new" or "difficult" for users, and which words can be identified as "learned" or "easy to learn" based on the user's existing knowledge level. This personalized learning plan can significantly reduce the user's learning cost and improve learning efficiency and motivation.
[0057] By identifying and predicting words that are easy for users to recognize, the present invention can directly mark these words as learned, thereby reducing the total amount of learning that learners need to face; this method effectively shortens the learning path and makes the learning process more efficient;
[0058] The present invention is not limited to word learning, but can be extended to any type of knowledge learning in theory, as long as there are clear difficulty levels of knowledge points and individual differences. BRIEF DESCRIPTION OF THE DRAWINGS
[0059] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:
[0060] Figure 1 It is a schematic diagram of a module of a concise and simple word intelligent management system based on deep learning of the present invention;
[0061] Figure 2 This is a flow chart of Embodiment 1 of a concise and simple word intelligent management method based on deep learning of the present invention;
[0062] Figure 3 This is a flow chart of Example 2 of a concise simple word intelligent management method based on deep learning of the present invention. DETAILED DESCRIPTION
[0063] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0064] See also Figure 1 , the present invention provides a technical solution:
[0065] An intelligent management system for concise simple words based on deep learning, the system includes: a word database management module, a vocabulary test module, a user answer data analysis module, a to-be-learned word list generation and update module, and a learning record management and feedback module;
[0066] The word database management module is responsible for managing the word database, including operations such as word storage, update, and deletion; calculating the difficulty score of each word, and dividing the words into different word clusters according to the score, to provide support for subsequent vocabulary testing and generation of lists of words to be learned;
[0067] The vocabulary test module creates vocabulary test questions based on word clusters, pushes the vocabulary test questions to the client, records the user's answer data, and determines whether to continue or complete the test based on the user's selection and answer situation;
[0068] The user answer data analysis module analyzes the answer data uploaded by the user after completing the vocabulary test to obtain the user's vocabulary prediction result; based on the user's answer data, the user's mastery ratio of each word cluster is obtained to construct the user's vocabulary feature vector;
[0069] The module for generating and updating the list of words to be learned obtains the prediction result of the probability distribution of the words that the user has mastered based on the user vocabulary feature vector, generates the list of words to be learned according to the prediction result of the probability distribution of the words that the user has mastered, and updates the list of words to be learned every day according to the user's learning record;
[0070] The learning record management and feedback module records the user's learning records, including daily learning status, learning plan completion status, etc., and provides feedback and personalized learning path suggestions based on the user's learning status; it also supports users to delete the learning word list, thereby generating a personalized learning path.
[0071] The word database management module includes a database connection unit, a word difficulty score calculation unit and a word cluster generation unit;
[0072] The database connection unit is responsible for establishing a connection with the word database and performing data reading and writing operations; the word difficulty score calculation unit calculates the difficulty score of each word and stores the score result in the database; the word cluster generation unit divides the word into several vocabulary clusters according to the word difficulty score.
[0073] The vocabulary test module includes a test question generation unit and a user answer recording unit;
[0074] The test question generating unit randomly selects words from each vocabulary cluster to generate vocabulary test questions, and pushes the vocabulary test questions to the client according to the user's request when the user first starts the application; the user answer recording unit records the user's choices in the test in real time, and monitors the user's choices after the vocabulary test reaches a preset number.
[0075] The user answer data analysis module includes an answer data uploading unit, a user vocabulary prediction unit, and a user vocabulary feature vector construction unit;
[0076] The answer data uploading unit receives the user answer data uploaded by the client and stores it on the server for analysis; the user vocabulary prediction unit analyzes the user's answer data and predicts the user's vocabulary; the user vocabulary feature vector construction unit obtains the user's mastery ratio of each word cluster based on the user's answer data and constructs a user vocabulary feature vector.
[0077] The to-be-learned word list generation and update module includes a to-be-learned word list generation unit and a to-be-learned word list update unit;
[0078] The to-be-learned word list generating unit obtains the prediction result of the probability distribution of the words that the user has mastered based on the user vocabulary feature vector, and generates the to-be-learned word list according to the prediction result of the probability distribution of the words that the user has mastered; the to-be-learned word list updating unit updates the to-be-learned word list according to the user's learning record of the previous day;
[0079] The learning record management and feedback module includes a learning record management unit and a personalized learning path generation unit;
[0080] The learning record management unit manages and records the user's learning record; the personalized learning path generation unit outputs the updated list of words to be learned to the user, and the user deletes the list of words to be learned to obtain a personalized learning path.
[0081] A concise and simple word intelligent management method based on deep learning, the method comprises the following steps:
[0082] Step S100. The user starts the word learning application through the client for the first time, and the server creates a vocabulary test and generates several test questions; the generated vocabulary test questions are pushed to the client, and the user takes the vocabulary test;
[0083] Step S200. The user completes the vocabulary test, and the client uploads the user's answer data to the server, which analyzes the user's answer data to obtain the user's vocabulary prediction result;
[0084] Step S300. Based on the analysis of the user's answer data by the server, a user vocabulary feature vector is constructed; based on the user vocabulary feature vector, the server obtains the probability distribution of the words that the user has mastered through a deep learning model based on a self-attention mechanism;
[0085] Step S400. The server generates a list of words to be learned based on the predicted results of the probability distribution of the words that the user has mastered; every other day, the list of words to be learned is updated based on the user's learning record of the previous day, thereby obtaining a personalized learning path.
[0086] Step S100 includes:
[0087] S101. The user starts the word learning application through the client for the first time, and the server creates a vocabulary test and generates several test questions; the process of creating a vocabulary test is as follows:
[0088] Get a word database and calculate the difficulty score of each word in the word database. The specific calculation formula for the difficulty score of a word is:
[0089] D=w1*L+w2*(1 / m)*(Σ i∈[1,m] F i )+w3*B,
[0090] Where D represents the difficulty score of the word; w1, w2 and w3 are all weight factors, and w1+W2+w3=1; L represents the length of the word; Fi represents the frequency of the word in the i-th test corpus; m represents the number of tests; B represents the difficulty level of the test with the most frequent word; according to the difficulty score of the word, the words are divided into several vocabulary clusters, and several words are randomly selected from each vocabulary cluster, and test questions are generated based on the selected words;
[0091] S102. The generated test questions are sent to the client in the form of a data packet. The client displays the test questions to the user in order, and the user selects the correct Chinese interpretation for each word displayed; the server records the user's selection in the test in real time. When the vocabulary test reaches a preset number, the user chooses to continue or complete the test.
[0092] Step S200 includes:
[0093] After the user completes the vocabulary test, the client uploads the user's answer data to the server; the server calculates the user data based on word cluster statistics to obtain the user's mastery of words in different word clusters; based on the user's mastery of words in different word clusters, the user's vocabulary S is predicted. The specific calculation formula for the user's vocabulary S is:
[0094] S=Σ c∈Clusters (int(Rc*Vc)),
[0095] Rc=[max(0.3*Kc-Uc)] / (3*Tc),
[0096] Among them, Kc represents the number of known words in word cluster c for the user, Uc represents the number of unknown words in word cluster c for the user, Tc represents the total number of words in word cluster c, and Tc=Kc+Uc+Cc, Cc represents the number of uncertain words in word cluster c for the user; Rc represents the mastery rate of word cluster c, Vc represents the length of the word list in word cluster c; Clusters represents the different word clusters involved in the user's vocabulary test, and each word cluster contains a group of related words.
[0097] The user's vocabulary mastery is assessed through an initial vocabulary test conducted on the client, which involves steps such as creating and pushing the test to the client, recording user selections in real time, and automatically terminating the test based on user responses. The core of this is the ability to dynamically adjust the test length and difficulty based on the user's actual performance until the predetermined confidence rate standard is reached, accurately assessing the user's vocabulary level.
[0098] Step S300 includes:
[0099] S301. For each user, based on the user's answer data, obtain the user's mastery ratio in each word cluster, and construct an n-dimensional feature vector, where n is equal to the number of word clusters; associate each user's n-dimensional feature vector with the corresponding user ID, thereby forming the user vocabulary feature vector V ID , and V ID ={v1,v2,...,vn}, where v1 represents the user's accuracy rate for the first word cluster, v2 represents the user's accuracy rate for the second word cluster, and so on, and vn represents the user's accuracy rate for the nth word cluster;
[0100] Based on the analysis of user test results, their vocabulary level is predicted and this information is converted into vocabulary feature vectors. This involves not only data analysis methods, but also how to effectively store these feature vectors in the server for subsequent customization of learning content. This is a key point in the technical solution because it is directly related to the accuracy and effectiveness of the subsequent personalized learning path.
[0101] S302. The user vocabulary feature vector V ID Perform weighted summation to obtain context coding K. The specific calculation formula of context coding K is:
[0102] K=Σ j∈[1,n] α j V j ,
[0103] Among them, α j represents the weight of the j-th word cluster, V j represents the accuracy of the jth word cluster, and j is a positive integer from 1 to n; j The specific calculation formula is:
[0104] α j =exp(e j ) / [Σ j∈[1,n] exp(e j )],
[0105] e j =ReLU(WqV j ) T (WxV ID ),
[0106] Among them, e j represents the activation value of the jth word cluster; ReLU() represents the rectified linear unit function; Wq and Wx represent parameter matrices, which are used to obtain the attention score according to the feature vector of each word cluster in the user vocabulary feature vector; the context encoding K is used to obtain the prediction result Y of the probability distribution of the words that the user has mastered, and Y=softmax(W0K), where W0 represents the parameter matrix of the output layer; in the above formula, ReLU and softmax are used for the activation function of the neural network and the probability distribution prediction of the output layer, respectively.
[0107] Step S400 includes:
[0108] S401. For each user, the server generates a personalized list of words to be learned based on the predicted results of the probability distribution of the words that the user has mastered; and the generation process of the list of words to be learned is as follows:
[0109] According to the prediction result of the probability distribution of the words that the user has mastered, each word in each word cluster is arranged in descending order according to the probability of the word that the user has mastered, and the words with probability equal to 1 are screened out, and the words with probability equal to 1 are marked as mastered words, and the words with probability less than 0.5 are marked as unmastered words, and the remaining words are marked as words to be consolidated; the marked words are allocated according to the user's learning time and plan, so as to generate a list of words to be learned;
[0110] S402. Every other day, the learning record of the user on the previous day is obtained, and a new user vocabulary feature vector is obtained based on the user's learning record on the previous day. The server passes the new user vocabulary feature vector through a deep learning model based on a self-attention mechanism, thereby obtaining a prediction result of the probability distribution of the new words that the user has mastered, and updates the list of words to be learned based on the prediction result of the probability distribution of the new words that the user has mastered, thereby obtaining a personalized learning path; and the user can delete the latest list of words to be learned output by the client, thereby reducing the time wasted on repeatedly learning known vocabulary and allowing the user to focus on learning new vocabulary.
[0111] During daily learning, the system uses the user's feedback on their mastery of new words to update the user's vocabulary level and adjust subsequent learning content accordingly. This process requires not only real-time data collection and processing, but also the ability to accurately analyze data to ensure that the learning content can accurately match the user's actual needs.
[0112] In this embodiment 1, Figure 2 As shown;
[0113] 1. Initial vocabulary evaluation and model training
[0114] 1.1. Trigger the creation of a vocabulary test by launching the word learning application through the client;
[0115] 1.2. Determine the difficulty level of each word and divide the words into several clusters according to the difficulty level;
[0116] 1.3. Randomly select words from each vocabulary cluster and generate test questions;
[0117] 1.4. The user takes a vocabulary test through the client and selects the correct Chinese meaning of each word;
[0118] 1.5. After the user completes the test, the client uploads the answer data to the server;
[0119] 1.6. The server counts the mastery of words in each word cluster based on the user data and constructs a user vocabulary feature vector;
[0120] 1.7. The server stores the user feature vector and uses the deep learning model to predict the words that the user has already mastered.
[0121] 2. New user vocabulary level assessment and word reduction
[0122] 2.1. When the user decides to start a new word learning task, the system generates a set of new words to be learned;
[0123] 2.2. During the learning process, users indicate their level of understanding of each word;
[0124] 2.3. After the learning task is completed, the server updates the statistics of the mastery of new words based on user feedback and updates the vocabulary feature vector;
[0125] 2.4. The system predicts the words that the user has mastered based on the updated vocabulary level and removes these words from the list of words to be learned;
[0126] 3. Update vocabulary level and delete words to be learned during daily learning
[0127] 3.1. The user continues to provide feedback on the degree of recognition of new words in subsequent learning;
[0128] 3.2. The server updates the vocabulary feature vector based on user feedback, predicts the words the user has mastered, and automatically trims the list to be learned;
[0129] 3.3. Repeat the above process to continuously update the user vocabulary knowledge base and provide a coherent and adaptive learning experience.
[0130] In this embodiment 2, Figure 3 As shown;
[0131] 1. This embodiment relates to a method for a user to learn each group of words in subsequent use, which is intended to further enhance the user's learning experience and optimize their vocabulary learning path.
[0132] 2. When the user decides to start a new word learning task, the system will generate and provide a set of new words to be learned.
[0133] 3. When users are learning words, they will be asked to provide feedback on their level of recognition of each word, that is, to indicate whether they know or do not know the new words they have learned.
[0134] 4. After the learning task is completed, the server will perform statistical calculations on the mastery of new words based on the user's feedback, obtain the statistical data changes in different word clusters, and update them to the vocabulary feature vector to ensure that the system has an accurate understanding of the user's level of understanding.
[0135] 5. Based on the updated vocabulary level, the system will further predict which words to be learned are actually already mastered by the user. Through intelligent prediction, the system can automatically remove these already mastered words from the user's list of words to be learned.
[0136] 6. This intelligent tailoring method ensures that users can focus on the words that have not yet been mastered in each learning task, thereby improving the overall learning efficiency and quality. By dynamically updating the user's vocabulary knowledge base, the system provides users with a coherent and adaptive learning experience.
[0137] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.
[0138] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art can still modify the technical solutions described in the aforementioned embodiments or replace some of the technical features therein by equivalents. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.
Claims
1. A concise and intelligent word management method based on deep learning, characterized by: The method comprises the following steps: Step S100. The user starts the word learning application through the client for the first time, and the server creates a vocabulary test and generates several test questions; the generated vocabulary test questions are pushed to the client, and the user takes the vocabulary test; Step S200. The user completes the vocabulary test, and the client uploads the user's answer data to the server, which analyzes the user's answer data to obtain the user's vocabulary prediction result; Step S300. Based on the analysis of the user's answer data by the server, a user vocabulary feature vector is constructed; based on the user vocabulary feature vector, the server obtains the probability distribution of the words that the user has mastered through a deep learning model based on a self-attention mechanism; Step S400. The server generates a list of words to be learned based on the prediction results of the probability distribution of the words that the user has mastered; every other day, the list of words to be learned is updated based on the user's learning record of the previous day, so as to obtain a personalized learning path; The step S300 includes: S301. For each user, based on the user's answer data, obtain the user's mastery ratio in each word cluster, and construct an n-dimensional feature vector, where n is equal to the number of word clusters; associate each user's n-dimensional feature vector with the corresponding user ID, thereby forming the user vocabulary feature vector V ID , and V ID ={v1,v2,...,vn}, where v1 represents the user's accuracy rate for the first word cluster, v2 represents the user's accuracy rate for the second word cluster, and so on, and vn represents the user's accuracy rate for the nth word cluster; S302. The user vocabulary feature vector V ID Perform weighted summation to obtain context coding K. The specific calculation formula of context coding K is: K=S j∈[1,n] a j V j , Among them, α j represents the weight of the j-th word cluster, V j represents the accuracy of the jth word cluster, and j is a positive integer from 1 to n; j The specific calculation formula is: α j =exp(e j ) / [Σ j∈[1,n] exp(e j )], it is j =ReLU(WqV j ) T (WxV ID ), Among them, e j Represents the activation value of the jth word cluster; ReLU() represents the rectified linear unit function; Wq and Wx represent parameter matrices, which are used to obtain the attention score based on the feature vector of each word cluster in the user's vocabulary feature vector; the context encoding K is used to obtain the predicted result Y of the probability distribution of the words that the user has mastered, and Y = softmax(W0K), where W0 represents the parameter matrix of the output layer.
2. According to claim 1, a concise and intelligent management method for simple words based on deep learning is characterized by: The step S100 includes: S101. The user starts the word learning application through the client for the first time, and the server creates a vocabulary test and generates several test questions; the process of creating a vocabulary test is as follows: Get a word database and calculate the difficulty score of each word in the word database. The specific calculation formula for the difficulty score of a word is: D=w1*L+w2*(1 / m)*(Σ i∈[1,m] F i )+w3*B, Where D represents the difficulty score of the word; w1, w2 and w3 are all weight factors, and w1+w2+w3=1; L represents the length of the word; Fi represents the frequency of the word in the i-th test corpus; m represents the number of tests; B represents the difficulty level of the test with the most frequent word; according to the difficulty score of the word, the words are divided into several vocabulary clusters, and several words are randomly selected from each vocabulary cluster, and test questions are generated according to the selected words; S102. The generated test questions are sent to the client in the form of a data packet. The client displays the test questions to the user in order, and the user selects the correct Chinese interpretation for each word displayed; the server records the user's selection in the test in real time. When the vocabulary test reaches a preset number, the user chooses to continue or complete the test.
3. The method for intelligent management of concise words based on deep learning according to claim 2 is characterized by: The step S200 includes: After the user completes the vocabulary test, the client uploads the user's answer data to the server; the server calculates the user data based on word cluster statistics to obtain the user's mastery of words in different word clusters; based on the user's mastery of words in different word clusters, the user's vocabulary S is predicted. The specific calculation formula for the user's vocabulary S is: S=Σ c∈Clusters (int(Rc*Vc)), Rc=[max(0.3*Kc-Uc)] / (3*Tc), Among them, Kc represents the number of known words in word cluster c for the user, Uc represents the number of unknown words in word cluster c for the user, Tc represents the total number of words in word cluster c, and Tc = Kc + Uc + Cc, Cc represents the number of uncertain words in word cluster c for the user; Rc represents the mastery rate of word cluster c, Vc represents the length of the word list in word cluster c; Clusters represents the different word clusters involved in the user's vocabulary test, and each word cluster contains a group of related words.
4. The method for intelligent management of concise words based on deep learning according to claim 1, characterized in that: The step S400 includes: S401. For each user, the server generates a personalized list of words to be learned based on the predicted results of the probability distribution of the words that the user has mastered; and the generation process of the list of words to be learned is as follows: According to the prediction result of the probability distribution of the words that the user has mastered, each word in each word cluster is arranged in descending order according to the probability of the word that the user has mastered, and the words with probability equal to 1 are screened out, and the words with probability equal to 1 are marked as mastered words, and the words with probability less than 0.5 are marked as unmastered words, and the remaining words are marked as words to be consolidated; the marked words are allocated according to the user's learning time and plan, so as to generate a list of words to be learned; S402. Every other day, obtain the user's learning record of the previous day, and obtain a new user vocabulary feature vector based on the user's learning record of the previous day. The server passes the new user vocabulary feature vector through a deep learning model based on the self-attention mechanism to obtain a new prediction result of the probability distribution of words that the user has mastered. According to the new prediction result of the probability distribution of words that the user has mastered, the list of words to be learned is updated to obtain a personalized learning path.
5. A concise and simple word intelligent management system based on deep learning, applied to a concise and simple word intelligent management method based on deep learning as claimed in any one of claims 1 to 4, characterized in that: The system includes: a word database management module, a vocabulary test module, a user answer data analysis module, a to-be-learned word list generation and update module, and a learning record management and feedback module; The word database management module is responsible for managing the word database, calculating the difficulty score of each word, and dividing the words into different word clusters according to the score; The vocabulary test module creates vocabulary test questions based on word clusters, pushes the vocabulary test questions to the client, records the user's answer data, and determines whether to continue or complete the test based on the user's selection and answer situation; The user answer data analysis module analyzes the answer data uploaded by the user after completing the vocabulary test to obtain the user vocabulary prediction result; obtains the user's mastery ratio of each word cluster based on the user's answer data, and constructs the user vocabulary feature vector; The to-be-learned word list generation and update module obtains a prediction result of the probability distribution of the words that the user has mastered based on the user vocabulary feature vector, generates a to-be-learned word list according to the prediction result of the probability distribution of the words that the user has mastered, and updates the to-be-learned word list every day according to the user's learning record; The learning record management and feedback module records the user's learning record and provides feedback and personalized learning path suggestions based on the user's learning situation; it also supports the user to delete the learning word list, thereby generating a personalized learning path.
6. The deep learning-based concise word intelligent management system according to claim 5 is characterized by: The word database management module includes a database connection unit, a word difficulty score calculation unit and a word cluster generation unit; The database connection unit is responsible for establishing a connection with the word database and performing data reading and writing operations; the word difficulty score calculation unit calculates the difficulty score of each word and stores the score result in the database; the word cluster generation unit divides the word into several vocabulary clusters according to the difficulty score of the word.
7. The deep learning-based concise word intelligent management system according to claim 5 is characterized by: The vocabulary test module includes a test question generating unit and a user answer recording unit; The test question generating unit randomly selects words from each vocabulary cluster to generate a vocabulary test question, and pushes the vocabulary test question to the client according to the user's request when the user starts the application for the first time; The user answer recording unit records the user's choices in the test in real time, and monitors the user's choices after the vocabulary test reaches a preset number.
8. The deep learning-based concise word intelligent management system according to claim 5, characterized in that: The user answer data analysis module includes an answer data uploading unit, a user vocabulary prediction unit and a user vocabulary feature vector construction unit; The answer data uploading unit receives the user answer data uploaded by the client and stores it on the server for analysis; the user vocabulary prediction unit analyzes the user's answer data and predicts the user's vocabulary; the user vocabulary feature vector construction unit obtains the user's mastery ratio of each word cluster based on the user's answer data and constructs a user vocabulary feature vector.
9. The deep learning-based concise word intelligent management system according to claim 5, characterized in that: The to-be-learned word list generation and update module includes a to-be-learned word list generation unit and a to-be-learned word list update unit; The to-be-learned word list generating unit obtains a prediction result of the probability distribution of the words that the user has mastered based on the user vocabulary feature vector, and generates a to-be-learned word list according to the prediction result of the probability distribution of the words that the user has mastered; the to-be-learned word list updating unit updates the to-be-learned word list according to the user's learning record of the previous day; The learning record management and feedback module includes a learning record management unit and a personalized learning path generation unit; The learning record management unit manages and records the user's learning record; the personalized learning path generation unit outputs the updated list of words to be learned to the user, and the user deletes the list of words to be learned to obtain a personalized learning path.
