Method and device for improving word learning efficiency of middle and primary school students

Abstract

A method and device for improving word learning efficiency of middle and primary school students relates to the technical field of recommendation learning, and comprises the following steps: S1, defining feature vectors of words and a word list of learning materials; s2, establishing feature vectors of all words in the English learning textbook, initializing a word recommendation value dictionary and performing learning material storage; s3, dynamically updating the word feature vectors and the recommendation value dictionary according to the expressions of the students in different learning types in the learning process; s4, screening out alternative learning materials, calculating recommendation values of the alternative learning materials through a word dictionary and a word recommendation value dictionary of the alternative learning materials, and pushing the learning material with the highest recommendation value to the student; the problem that traditional word learning lacks personalized dynamic recommendation and is difficult to adapt to multi-modal learning scenes and individual memory attenuation laws is solved, dynamic feature engineering is combined with multi-scene recommendation, visual closed-loop optimization of long-term learning effects is realized, and word learning efficiency is effectively improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of recommendation learning, in particular to a method and device for improving the word learning efficiency of primary and secondary school students. BACKGROUND

[0002] An important link for primary and secondary school students to learn English is to memorize words. After mastering words, a solid foundation is laid for oral English, dictation, listening, reading and writing. Then, students often spend a lot of time on repeatedly memorizing words that they have already mastered. Relatively speaking, for words that are not well mastered, the learning content of the students may not involve enough of these words, and they do not spend enough time learning them.

[0003] At present, it is very common for students to purchase educational electronic products such as student tablets to learn English. In the past, educational electronic products mainly judged the mastery of words by word tests, and recommended word learning content based on this. The recommended way is often to list the words that have not been mastered for relearning, and does not identify words that need to be strengthened from oral English, dictation, listening, reading and writing, resulting in inaccurate and comprehensive mastery of students' words, and it is difficult to achieve accurate push of oral English, dictation, listening, reading and writing materials.

[0004] Therefore, there is an urgent need for a method that can fully consider the difficulty, importance, frequency of occurrence, mastery level and forgetting law of words and other factors, provide personalized learning recommendations for students, meet the learning needs of different students, and effectively improve the word learning efficiency of primary and secondary school students. SUMMARY

[0005] In order to overcome the deficiencies in the background art, the present application provides a method and device for improving the word learning efficiency of primary and secondary school students.

[0006] To achieve the above-mentioned application purposes, the present application adopts the following technical solutions:

[0007] In a first aspect, the present application provides a method for improving the word learning efficiency of primary and secondary school students, comprising the following steps:

[0008] S1, defining a feature vector of a word and a word list of learning materials; the feature vector includes the number of phonemes of the word, the importance of the word, the inverse of the cumulative number of occurrences of the word, the inverse of the last learning mastery level, and the time stamp of the last learning;

[0009] S2, establishing a feature vector of all words in the English learning materials according to the outline requirements and the content of the teaching materials, then calculating the word recommendation value according to the feature vector, initializing the word recommendation value dictionary and performing learning material warehousing; the learning materials contain the corresponding word dictionary;

[0010] S3, dynamically updating the word feature vector and the recommended value dictionary according to the performance of the student in different learning types in the learning process; the learning types include word learning, oral, dictation, listening, reading and writing;

[0011] S4, filtering out the candidate learning materials according to the grade information and the learning type, calculating the recommended value of the candidate learning materials through the word dictionary of the candidate learning materials and the word recommended value dictionary, and pushing the learning material with the highest recommended value of the candidate learning materials to the student.

[0012] Specifically, the feature vector in step S2 calculates the word recommended value WR through formula (1), and the formula (1) is as follows:

[0013] WR=W1*v1+W2*v2+W3*v4+W4*v4+W5*v5’ (1)

[0014] Wherein, v1 is the number of phonemes of the word, W1 is the weight of the number of phonemes of the word; v2 is the word importance, W2 is the word importance weight; v3 is the reciprocal of the cumulative number of occurrences of the word, W3 is the reciprocal weight of the cumulative number of occurrences of the word; v4 is the reciprocal of the mastery degree of the last learning, W4 is the reciprocal weight of the mastery degree of the last learning; v5' is the forgetting coefficient obtained by converting the last learning time stamp v5, W5 is the forgetting coefficient weight;

[0015] Specifically, the calculation formula of the forgetting coefficient is as shown in formula (2):

[0016] Fc=1-R (2)

[0017] Wherein, Fc is the forgetting coefficient; R is the memory retention rate of Ebbinghaus, that is: R=e^(-t / S), wherein e is the base number of natural logarithm, t is the forgetting time, unit: day, t=(ts-v5) / 86400; S is the relative memory strength; ts is the current time stamp, v5 is the last learning time stamp in the feature vector.

[0018] Specifically, the process of step S2 includes:

[0019] (1) generating a unique identifier material_id of the material by using self-increment ID, UUID or hash algorithm;

[0020] (2) integrating the material content, word dictionary, grade, learning type and ID into a structured object;

[0021] (3) saving the structured object into the database through the database, which is convenient for subsequent use.

[0022] Specifically, step S3 specifically includes:

[0023] The reciprocal of the cumulative occurrence number of the word, the reciprocal of the mastery level at the last learning, and the time stamp of the last learning of the corresponding feature vector are updated during the word learning, word learning follow-up reading and spelling process, and the word recommendation value is recalculated, and then the word recommendation value dictionary is updated; Specifically:

[0024] The reciprocal of the cumulative occurrence number of the word is obtained from the reciprocal of the reciprocal of the cumulative occurrence number of the word in the word feature vector, and the latest reciprocal of the cumulative occurrence number of the word is obtained by taking the reciprocal of the cumulative occurrence number of the word after adding 1;

[0025] The time stamp of the last learning is updated to the time stamp when the current word is learned;

[0026] The reciprocal of the latest mastery level v4' of the word is calculated according to formula (5)-(6) according to the word follow-up reading accuracy score Fl, the spelling accuracy score Sp and the time efficiency score Te, and then the reciprocal of the mastery level at the last learning of the word feature vector is replaced; Finally, the word recommendation value WR of the word is recalculated according to the updated word feature vector by using the calculation formula of the word recommendation value, and then the word recommendation value dictionary is updated; Formula (5)-(6) is as follows:

[0027] Mn=[Fl*a+Sp*b+Te*c] (5)

[0028] Wherein, Mn is the current mastery level of the word, a is the follow-up reading accuracy score weight, b is the spelling accuracy score weight, and c is the time efficiency score weight; a+b+c=1;

[0029] The calculation formula of the reciprocal of the latest mastery level v4' is shown in formula (6):

[0030] v4’=R*v4+(1-R)*1 / Mn (6)

[0031] Wherein, v4' is the reciprocal of the latest mastery level, v4 is the reciprocal of the mastery level at the last learning in the word feature vector, R is the memory retention rate of Ebbinghaus, that is: R=e^(-t / S); Wherein e is the base number of natural logarithm, t is the forgetting time, unit: day, t=(ts-v5) / 86400; S is the relative memory strength, the value is 1.84; ts is the current time stamp, and v5 is the time stamp of the last learning in the feature vector.

[0032] Specifically, the word follow-up reading accuracy score Fl is obtained by scoring the voice of the follow-up reading word through the third-party voice recognition technology;

[0033] The spelling accuracy score Sp is determined according to formula (3):

[0034] Sp = [(correctNum / totalNum)*100] (3)

[0035] Wherein, correctNum is the number of correct spelling letters, totalNum is the total number of letters in a word, and [] is the rounding symbol.

[0036] The time efficiency score Te is determined according to the time for spelling a correct word, as shown in equation (4).

[0037] Te = Min[(T / t)*100, 100] (4)

[0038] Wherein, Te is the time efficiency score, T = n*T1, T1 is the experienced time length for inputting an English letter on an electronic device, in ms, n represents the number of letters in a word, and t is the actual time length spent by the user for inputting a word, in ms.

[0039] Specifically, step S3 further comprises:

[0040] The oral learning is to identify the pronunciation score of each word in the read sentence according to the speech recognition technology, calculate the current mastery degree of each word according to the pronunciation score of each word, calculate the inverse of the latest mastery degree of the word according to equation (6), and update the inverse of the cumulative number of times of the word, the inverse of the mastery degree at the last learning, and the time stamp of the last learning of the corresponding feature vector, and then recalculate the word recommendation value, and update the word recommendation value dictionary.

[0041] Specifically, step S3 further comprises:

[0042] The dictation learning is to determine the current mastery degree of the word in different ways according to the spelling accuracy score of the word, calculate the inverse of the latest mastery degree of the word through equation (6) according to the current mastery degree of the word, then update the inverse of the cumulative number of times of the word, the inverse of the mastery degree at the last learning, and the time stamp of the last learning of the corresponding feature vector, and recalculate the word recommendation value, and then update the word recommendation value dictionary.

[0043] When the spelling of the word is completely correct, the time efficiency score Te is calculated by equation (4), and is taken as the current mastery degree Mn of the word, i.e. Mn = Te.

[0044] When the spelling of the word is not completely correct, the current mastery degree Mn of the word is calculated by equation (7):

[0045] Mn = U1*Sp + (1-U1)*Te (7)

[0046] Wherein, U1 is an experience value, which is a floating point number between 0 and 1.

[0047] Specifically, step S3 further comprises:

[0048] For writing learning, the reciprocal of the mastery degree at the last learning time of the word is calculated when the word in the writing content is in the word recommendation value dictionary and the spelling of the word is correct, then the reciprocal of the word cumulative occurrence frequency of the corresponding feature vector, the reciprocal of the mastery degree at the last learning time and the time stamp of the last learning are updated, and the word recommendation value is recalculated, and then the word recommendation value dictionary is updated;

[0049] When the writing is completed, all words are obtained according to the writing content, and all words are traversed: if the word is in the word recommendation value dictionary, the calculation time efficiency score Te is calculated through formula (4) according to the time length spent by the word input, and the calculation time efficiency score Te is taken as the current mastery degree Mn of the word; the current mastery degree Mn of the word obtained by calculation is used to calculate the reciprocal v4' of the latest mastery degree of the word according to formula (6), and then the reciprocal of the latest mastery degree is used to replace the reciprocal of the mastery degree at the last learning time of the word feature vector; finally, the word recommendation value of the word is recalculated according to the updated word feature vector by using the calculation formula of the word recommendation value, and then the word recommendation value dictionary is updated;

[0050] Specifically, step S3 further comprises:

[0051] For listening / reading, if the listening / reading answer is completely correct, all words in the word dictionary of the listening / reading material are traversed, if the traversed word is in the recommendation value dictionary, the reciprocal of the word cumulative occurrence frequency of the corresponding feature vector of the word and the time stamp of the last learning are updated, and the current mastery degree Mn of the word is set to 100; then the reciprocal v4' of the latest mastery degree of the word is calculated according to formula (6) through the current mastery degree Mn of the word;

[0052] If the listening / reading answer is not completely correct, all words in the word dictionary of the listening / reading material are traversed to find the n words with the highest word recommendation value in the word recommendation value dictionary as the words not understood, and the reciprocal of the word cumulative occurrence frequency of the corresponding feature vector of the words not understood and the time stamp of the last learning are updated; the value of n is determined according to the length of the word dictionary of the listening material; the reciprocal v4' of the latest mastery degree is calculated according to formula (8):

[0053] v4' = R1 * R * v3 + (1 - R1) * (1 - R) * 1 / Mn (8)

[0054] Wherein, v4' is the reciprocal of the latest mastery degree, v4 is the reciprocal of the mastery degree at the last learning in the word feature vector, R is the memory retention rate of Ebbinghaus, Mn is the current mastery degree of the word; the value of Mn is determined by multiplying the proportion of the final score of the listening to the total score by 100; R1 is the memory influence coefficient;

[0055] Then, the inverse of the latest mastery degree v4' replaces the inverse of the mastery degree at the last learning of the word feature vector; finally, the word recommendation value of the word is recalculated according to the updated word feature vector by using the calculation formula of the word recommendation value, and then the word recommendation value dictionary is updated.

[0056] Specifically, step S4 specifically includes the following steps:

[0057] S41, when recommending learning materials, first, according to the grade information and learning type of the student, the learning materials suitable for the age of the student are selected as the candidate learning materials;

[0058] S42, according to all the words in the word dictionary of the candidate learning materials, the word recommendation value dictionary is traversed, the m words with the highest word recommendation value are found out, and the material recommendation dictionary is generated;

[0059] S43, if the learning type is oral, dictation, listening, reading or writing, the total recommendation value of the learning materials is calculated according to the material recommendation dictionary and formula (9), and the candidate learning materials with the highest total recommendation value of the learning materials are pushed to the student;

[0060]

[0061] Wherein, MLR is the total recommendation value of the learning materials, count i is the number of times the i-th word in the material recommendation dictionary appears in the learning materials, i is the word appearance frequency ranking; i=1 represents the highest ranking; WR i is the word recommendation value of the i-th word in the material recommendation dictionary;

[0062] S44, if the learning type is word learning, the total recommendation value of the learning materials is calculated according to the material recommendation dictionary and formula (9), and the words in the word dictionary of the candidate learning materials with the highest total recommendation value of the learning materials are sorted according to their recommendation value in the word recommendation value dictionary from high to low and pushed to the student.

[0063] Specifically, W1+W2+W3+W4+Wn=1 in formula (1).

[0064] Specifically, W1 is 0.15; W2 is 0.35; W3 is 0.1; W4 is 0.25; and W5 is 0.15.

[0065] Specifically, the value of U1 in formula (7) is 0.75.

[0066] Specifically, the memory influence coefficient R1 in formula (8) is 0.85.

[0067] In a second aspect, the present application provides a device for improving the efficiency of word learning of students in primary and secondary schools, based on the method for improving the efficiency of word learning of students in primary and secondary schools described in the first aspect, comprising the following units:

[0068] An initialization unit is configured to define a feature vector of a word and a word list of learning materials; the feature vector includes the number of phonemes of the word, the importance of the word, the inverse of the cumulative number of occurrences of the word, the inverse of the last learning mastery level, and the timestamp of the last learning;

[0069] A construction unit is configured to establish a feature vector of all words in an English learning material according to the outline requirements and the content of the teaching material, then calculate a word recommendation value according to the feature vector, initialize a word recommendation value dictionary, and perform learning material warehousing; the learning material contains a corresponding word dictionary;

[0070] An updating unit is configured to dynamically update the feature vector of the word and the recommendation value dictionary according to the performance of the student in different learning types in the learning process; the learning types include word learning, oral English, dictation, listening, reading, and writing;

[0071] A recommendation unit is configured to filter out candidate learning materials according to the grade information and the learning type, calculate the recommendation value of the candidate learning materials by the word dictionary of the candidate learning materials and the word recommendation value dictionary, and push the learning material with the highest recommendation value of the candidate learning materials to the student.

[0072] In a third aspect, the present application further provides an electronic device, comprising a processor, a memory, a communication interface, and one or more programs, the one or more programs being stored in the memory and configured to be executed by the processor, the program comprising instructions for performing the steps in the method described in the first aspect.

[0073] The application provides a method and device for improving the word learning efficiency of primary and secondary school students, and comprises the following steps: S1, defining the feature vector of a word and the word list of learning materials; S2, establishing the feature vector of all words in the English learning materials according to the outline requirements and the content of the teaching materials, then calculating the word recommendation value according to the feature vector, initializing the word recommendation value dictionary and carrying out the learning material warehousing; S3, dynamically updating the word feature vector and the recommendation value dictionary according to the performance of students in different learning types in the learning process; S4, screening the selected learning materials according to the grade information and the learning type, calculating the selected learning material recommendation value through the word dictionary and the word recommendation value dictionary of the selected learning materials, and pushing the learning material with the highest selected learning material recommendation value to the students; the application solves the problems of low traditional word learning efficiency, lack of personalized dynamic recommendation mechanism and difficulty in adapting to the multi-modal learning scene and individual memory decay law, and identifies the words to be strengthened from the usual word learning, oral English, dictation, listening, reading and writing, realizes the long-term learning effect visual closed-loop optimization through the combination of dynamic feature engineering and multi-scene recommendation, and effectively improves the efficiency of word learning.

[0074] Further, the traditional word learning adopts a fixed order or random repetition, and the dynamic changes of word difficulty and student mastery degree are not considered, resulting in inefficient repetition or omission of key words, the embodiment quantifies the word learning priority through the feature vector (such as the number of phonemes, importance, and cumulative appearance frequency reciprocal), adjusts the recommendation value in combination with the dynamic performance of students, and realizes accurate learning; through the dynamic recommendation value dictionary, the words with low student mastery degree, high importance and in line with the current learning type are preferentially pushed, invalid repetition is reduced, and the time to reach the standard is shortened; the multi-dimensional feature vector (such as the timestamp decay memory curve and the cumulative appearance frequency reciprocal to avoid “high-frequency word monopoly”) replaces the subjective experience order, so that the learning path is more in line with the cognitive law.

[0075] Further, the traditional method lacks real-time feedback mechanism for student learning behavior, and cannot dynamically adjust the learning strategy, the embodiment dynamically updates the word feature vector (such as adjusting the “last mastery degree reciprocal” according to the dictation error rate) through step S3, realizes the continuous optimization of the learning path, combines the performance data of the oral English, dictation and other scenes, reversely optimizes the word recommendation strategy, and forms a “learning-evaluation-optimization” closed loop.

[0076] Further, through the timestamp and mastery degree reciprocal in the feature vector, the word memory decay curve can be tracked, and data support can be provided for subsequent review plan. BRIEF DESCRIPTION OF DRAWINGS

[0077] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed in the embodiments. Obviously, the drawings in the following description only constitute some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained from these drawings without creative effort.

[0078] Figure 1 is a method for improving the learning efficiency of words of primary and secondary school students according to an embodiment of the present application;

[0079] Figure 2 is a device for improving the learning efficiency of words of primary and secondary school students according to an embodiment of the present application;

[0080] Figure 3 is a device for improving the learning efficiency of words of primary and secondary school students according to an embodiment of the present application. DETAILED DESCRIPTION

[0081] The present application can be explained in detail through the following embodiments, and the purpose of the present application is to protect all technical improvements within the scope of the present application. In the description of the present application, it should be understood that if the terms "upper", "lower", "front", "rear", "left", "right" and the like indicate the orientation or positional relationship, they are only corresponding to the drawings of the present application, and are not intended to indicate or imply that the device or element must have a specific orientation.

[0082] Embodiment One

[0083] Reference Figure 1 The present embodiment provides a method for improving the learning efficiency of words of primary and secondary school students, comprising the following steps:

[0084] S1, defining a feature vector of a word and a word list of learning materials; the feature vector comprises the number of syllables of the word, the importance of the word, the inverse of the cumulative number of occurrences of the word, the inverse of the last learning mastery degree, and the time stamp of the last learning;

[0085] The definition of the feature vector of the word specifically comprises:

[0086] In the present embodiment, each word required by the primary and secondary school teaching materials is represented as a feature vector. In all feature values of the feature vector, except for the time stamp of the last learning, each feature value takes a floating point number from 0 to 1. The feature vector comprises the following features:

[0087] 1) The number of syllables of the word, representing the difficulty of the word, taking 0.1 for one syllable, 0.2 for two syllables, and so on;

[0088] The number of syllables of an English word depends on the number of vowels in the word and the pronunciation rules. In simple terms, each vowel or vowel combination usually forms a syllable. The number of syllables of an English word can be obtained by writing a corresponding code according to the above rules, for example, by directly obtaining the number of syllables of a word from a pre-stored pronunciation database (such as the CMU pronunciation dictionary), or by using an existing natural language processing (NLP) tool library to count the built-in syllables of an English word. Further, the number of syllables can be predicted by a trained model according to the spelling features (letter sequence, vowel-consonant combination, prefix / suffix, etc.) of the word. The present embodiment is not limited in this regard, and any method that can obtain the number of syllables of a word is within the scope of protection.

[0089] Preferably, the CMU pronunciation dictionary covers the vocabulary of primary and secondary school teaching materials, so that the present embodiment obtains the number of syllables of a word by calling the API of a pre-stored pronunciation database, which is Pronouncing API.

[0090] Pronouncing API is based on CMUPronouncingDictionary (CMU Pronunciation Dictionary), which is developed by Carnegie Mellon University and contains a large number of English words and their corresponding phonetic (phoneme) representations, covering phoneme annotation of more than 130,000 English words. Pronouncing API provides a series of functions such as obtaining word pronunciation, finding rhyming words, and calculating the number of syllables by parsing and indexing these phonetic data. CMU dictionary annotates the phoneme sequence of a word, and the number of syllables is determined by the number of vowel phonemes (Vowel) in the phoneme.

[0091] For example, the number of syllables of the word "example" is 3, and the number of syllables of the word "apple" is 2.

[0092] 2) Word importance, which is divided according to the grade of primary and secondary school vocabulary. In the present embodiment, the vocabulary of primary and secondary school teaching materials is divided into four categories, including primary and secondary school basic core vocabulary, primary and secondary school expansion commonly used vocabulary, primary and secondary school advanced vocabulary, and primary and secondary school non-essential vocabulary. According to the grade and experience value of 0.9 to 0.1, the greater the value, the more important the word.

[0093] In the present embodiment, the value of the primary and secondary school basic core vocabulary is 0.9, the value of the primary and secondary school expansion commonly used vocabulary is 0.7, the value of the primary and secondary school advanced vocabulary is 0.4, and the value of the primary and secondary school non-essential vocabulary is 0.1.

[0094] In one possible implementation, the vocabulary of primary and secondary school teaching materials can be divided into more or fewer types as needed, and the division can be made as needed. The importance of the divided words can also be adjusted as needed.

[0095] In another possible implementation, the scope of the learning material can also be extended and divided, such as introducing vocabulary related to college, IELTS, TOEFL, etc. The above is not limited in particular.

[0096] 3) The reciprocal of the cumulative number of occurrences of the word, which is used to measure the frequency of the word in the learning process. The more frequently a word appears, the more likely it is to have been mastered. At this time, the reciprocal of the cumulative number of occurrences is smaller. The initial value of this attribute of a word is the reciprocal of the cumulative number of occurrences of the word in the textbook. As the learning on the specific APP progresses, the reciprocal of the cumulative number of occurrences of the word will change in real time. The initial value of the reciprocal of the cumulative number of occurrences of the word is set to 0.

[0097] 4) The reciprocal of the last learning mastery level, which is valued from 1 to 100. The less a word is mastered, the smaller the value of the mastery level, and the larger the reciprocal of the mastery level. The initial value of the mastery level is 1. This embodiment divides the mastery level into four levels, which are:

[0098] 1-25: Not mastered (completely unfamiliar, unable to recall)

[0099] 26-50: Preliminary mastery (have a vague impression, need to be prompted to recall)

[0100] 51-75: Basic mastery (can recall independently, but not very skilled)

[0101] 76-100: Skilled mastery (accurate and rapid recall, flexible application).

[0102] 5) The timestamp of the last learning, which is the number of seconds from the current time to January 1, 1970, 00:00:00. The initial value of the reciprocal of the last learning mastery level is set to 0.

[0103] It is worth noting that the number of features of the feature vector in this embodiment is an integer greater than or equal to 5, which can be further expanded based on the above five feature vectors, which are not limited here.

[0104] In this way, a word word can be represented as a feature vector:

[0105] word=[v1,v2,v3,v4,v5,...vn];

[0106] Where n is the number of features, n is an integer greater than 5, v1 is the number of phonemes in the word, v2 is the importance of the word, v3 is the inverse of the cumulative number of occurrences of the word, v4 is the inverse of the degree of mastery of the last learning, v5 is the timestamp of the last learning,... vn represents other more feature vectors, except that the timestamp of the last learning belongs to the integer value, all other feature values (i.e. v1, v2, v3, v4, v6,... vn) are floating-point numbers between 0 and 1.

[0107] The vector format of multiple words is as follows:

[0108] word1 = [v11, v21, v31,..., vn1]

[0109] word2 = [v12, v22, v32,..., vn2]

[0110] Except for the first two feature values (the number of phonemes in the word and the importance of the word), other feature values change dynamically in the student learning process.

[0111] The word list WordDict of each learning material is defined as:

[0112] The word list WordDict is stored in a dictionary format, with the key name being the word and the value being the number of occurrences of the word in the article; the type of learning material includes oral, dictation, listening, reading and writing;

[0113] The definition of WordDict is: {"apple":5;"what":2;}, where "apple","what": refers to the word itself, and the number represents the cumulative number of occurrences of the word in the learning material.

[0114] Taking a reading type learning material as an example:

[0115] The title of the learning material is: An Apple Adventure;

[0116] The material content of the learning material is:

[0117] What a delicious apple! Emma picked a bright red apple from the tree. She wondered, What kind of apple is this? It smelled sweet and tastedcrisp. Emma shared the apple with her friend, saying, 'What a joy to enjoy nature's gift! 'Her friend replied, 'What a perfect day for an apple picnic! '

[0118] The vocabulary list of the learning materials is WordDict = {"what":4,"a":4,"delicious":1,"apple":4,"emma":2,"picked":1,"bright":1,"red":1,"from":1,"the":3,"tree":1,"she":2,"wondered":1,"kind":1,"is":1,"this":1,"it":1,"smelled":} 1,"sweet":1,"and":1,"tasted":1,"crisp":1,"shared":1,"with":1,"her":2,"friend":2,"saying":1,"jo y":1,"to":1,"enjoy":1,"natures":1,"gift":1,"replied":1,"perfect":1,"day":1,"for":1,"picnic":1};

[0119] S2. Based on the syllabus requirements and textbook content, establish feature vectors for all words in the English learning materials, then calculate word recommendation values ​​based on the feature vectors, initialize the word recommendation value dictionary, and input the learning materials into the database; the learning materials include the corresponding word dictionary;

[0120] The English learning materials mentioned are English learning materials for primary and secondary schools;

[0121] The syllabus requirements refer to the English teaching standards or guidance documents for primary and secondary schools formulated by national or local education authorities. They clearly define the English knowledge and skills students should master at different learning stages, including but not limited to vocabulary, grammar points, and the development goals for listening, speaking, reading, and writing abilities. The syllabus requirements are authoritative and instructive, serving as the direct basis for textbook compilation and teaching.

[0122] Textbook content refers to the content of specific primary and secondary school English textbooks. It directly reflects the syllabus requirements and is the main material for students to learn English. Textbook content typically includes texts, vocabulary dictionaries, exercises, and other sections.

[0123] The official websites of national and local education authorities usually publish the latest teaching syllabus and textbook information. Textbook publishers can also provide textbook catalogs, sample chapters, and syllabus-related resources.

[0124] When representing a word as a feature vector, the first three feature values—the number of syllables, word importance, and the reciprocal of the cumulative frequency of the word—are initially determined based entirely on the syllabus requirements and textbook content. Furthermore, the values ​​of the number of syllables and word importance remain unchanged after initialization. The initial values ​​of the other feature values ​​are set to 0 by default, since there is initially no feedback from student learning.

[0125] In this step, we calculate the recommendation value for all words in primary and secondary school English textbooks, and then initialize a recommendation value dictionary, WordRecommendDict, for all words. Each item in this dictionary specifies a word and its corresponding recommendation value. The higher the recommendation value, the more unfamiliar the student is with the word, and the more necessary it is to recommend learning materials containing that word. This is stored in the form of word-recommend value key-value pairs. The content of the recommendation value dictionary, WordRecommendDict, is similar to: {"apple":0.496,"what":0.323,}, where "apple" and "what" refer to the word itself; 0.496 and 0.323 are the recommendation values ​​corresponding to the word. The higher the recommendation value, the greater the probability of recommending learning content containing that word when recommending learning content in the future.

[0126] Specifically, the word recommendation value is calculated by weighting and summing the word's feature vector with the preset feature weight values;

[0127] As shown above, a word can be represented as a feature vector:

[0128] word=[v1,v2,v3,v4,v5,...vn];

[0129] The formula for calculating the word recommendation score (WR) is as follows:

[0130] WR=W1*v1+W2*v2+W3*v4+W4*v4+W5*v5'+...+Wn*vn

[0131] Where v5' is the forgetting coefficient obtained by converting v5 (the timestamp of the last learning). W1, W2...Wn represent the weights of v1, v2...vn in the word recommendation value calculation, representing the importance of v1, v2...vn in the calculation of word recommendation value. They are floating-point numbers between 0 and 1, and satisfy W1+W2+...+Wn=1.

[0132] In this embodiment, taking five features of a word vector as an example, the formula for calculating the word recommendation value WR is as shown in formula (1):

[0133] WR=W1*v1+W2*v2+W3*v4+W4*v4+W5*v5' (1)

[0134] In this context, v1 (number of syllables in the word) represents word difficulty; generally, more difficult words are harder to remember, so W1 can be slightly larger. In this example, the weight of W1 for the number of syllables in the word is 0.15. v2 (word importance) is also important; more important words should be recommended more, while less important words should be recommended less. The weight of W2 for word importance should be larger than W1; in this example, W2 is 0.35. v3 (reciprocal of the cumulative number of times a word appears) is relatively less important; in this example, the weight of W3 for the cumulative number of times a word appears is 0.1. v4 (reciprocal of the level of mastery at the time of the last learning) is more convincing in judging the degree of mastery of a word, so this weight should be relatively large; in this example, the weight of W4 for the reciprocal of the level of mastery at the time of the last learning is 0.25. v5' is the forgetting coefficient obtained by converting v5 (timestamp of the last learning), which is also relatively important; in this example, the weight of W5 for the forgetting coefficient is 0.15. The values ​​of W1, W2, ..., Wn can be fine-tuned based on actual test results.

[0135] The forgetting coefficient can be calculated using the timestamp of the last learning session.

[0136] The formula for calculating the forgetting coefficient is shown in formula (2):

[0137] Fc=1-R (2)

[0138] Where Fc is the forgetting coefficient; R is the Ebbinghaus forgetting curve, i.e., R = e^(-t / S), where e is the base of the natural logarithm (approximately 2.718), t is the forgetting time in days, t = (current timestamp - last learning timestamp) / 86400, where 86400 is the number of seconds in a day, and S is the relative memory strength, with a value of 1.84. Thus, the forgetting coefficient is a floating-point number between 0 and 1, with a time span t = (ts - v5) / 86400. A larger forgetting coefficient indicates a larger forgetting time span; ts is the current timestamp, and v5 is the last learning timestamp in the feature vector.

[0139] Here are some examples of words:

[0140] Example 1: Core vocabulary for primary and secondary school students: the

[0141] Word syllable count v1: 1 → 0.1

[0142] Word importance v2: 0.9

[0143] Countdown of cumulative occurrences v3: 200 occurrences in the textbook → 1 / 200 = 0.005

[0144] The last four counts of mastery level from the previous lesson: 0

[0145] Last learning timestamp v5: 0

[0146] Feature vector: [0.1, 0.9, 0.005, 0, 0]

[0147] The recommended value (WR) for the word "the" is: WR = 0.15 * 0.1 + 0.35 * 0.9 + 0.1 * 0.005 + 0 + 0 = 0.3305

[0148] Example 2: Commonly used vocabulary for primary and secondary school students: happy

[0149] Word syllable count v1: 2 → 0.2

[0150] Word importance v2: 0.7

[0151] Reverse count of cumulative occurrences v3: 30 occurrences in the textbook → 1 / 30 ≈ 0.033

[0152] The last four counts of mastery level from the previous lesson: 0

[0153] Last learning timestamp v5: 0

[0154] Eigenvector: [0.2, 0.7, 0.033, 0, 0]

[0155] The recommended value (WR) for the word "happy" is: WR = 0.15 * 0.2 + 0.35 * 0.7 + 0.1 * 0.033 + 0 + 0 = 0.2783

[0156] Example 3: Advanced vocabulary for primary and secondary school students: science

[0157] Word syllable count v1: 2 → 0.2

[0158] Word importance v2: 0.4

[0159] Reverse count of cumulative occurrences v3: Appears 8 times in the textbook → 1 / 8 = 0.125

[0160] The last four counts of mastery level from the previous lesson: 0

[0161] Last learning timestamp v5: 0

[0162] Eigenvector: [0.2, 0.4, 0.125, 0, 0]

[0163] The recommended value (WR) for the word "science" is: WR = 0.15 * 0.2 + 0.35 * 0.4 + 0.1 * 0.125 + 0 + 0 = 0.1825

[0164] Example 4: Non-essential vocabulary for primary and secondary schools: luxury

[0165] Word syllable count v1: 2 → 0.2

[0166] Word importance v2: 0.1

[0167] Reverse count of cumulative occurrences v3: Appears 3 times in the textbook → 1 / 3 ≈ 0.333

[0168] The last four counts of mastery level from the previous lesson: 0

[0169] Last learning timestamp v5: 0

[0170] Eigenvector: [0.2, 0.1, 0.333, 0, 0]

[0171] The recommended value (WR) for the word "luxury" is: WR = 0.15 * 0.2 + 0.35 * 0.1 + 0.1 * 0.333 + 0 + 0 = 0.0983

[0172] This step also includes the input of English learning materials into the database. When inputting learning materials, they will be categorized and stored according to grade level and learning type. Each learning material also comes with a corresponding WordDict dictionary for input. The learning types include vocabulary learning, speaking, dictation, listening comprehension, reading, and writing. The grade levels include: Grade 1, Grade 2, Grade 3, Grade 4, Grade 5, Grade 6, Grade 7 (Junior High 1), Grade 8 (Junior High 2), Grade 9 (Junior High 3), Grade 10, Grade 11, and Grade 12.

[0173] Taking reading materials as an example, when each piece of reading material is added to the database, in addition to the full text of the material itself, it also includes a WordDict dictionary, which contains all the words in the reading material and the frequency of each word.

[0174] The format of WordDict is similar to: {"apple":5;"what":2;}, where "apple" and "what" refer to the word itself, and the numbers represent the cumulative number of times the word appears in the learning content.

[0175] For example, let's take a reading material for third graders as an example:

[0176] The title of the reading material is: An Apple Adventure;

[0177] The content of the reading material is as follows:

[0178] What a delicious apple! Emma picked a bright red apple from the tree. She wondered, What kind of apple is this? It smelled sweet and tastedcrisp. Emma shared the apple with her friend, saying, 'What a joy to enjoy nature's gift! 'Her friend replied, 'What a perfect day for an apple picnic! '

[0179] The word dictionary of the reading material WordDict={"what":4,"a":4,"delicious":1,"apple":4,"emm a":2,"picked":1,"bright":1,"red":1,"from":1,"the":3,"tree":1,"she":2,"wondered":1,"kind":1,"is":1,"this":1,"it":1,"smelled":1,"sweet":1,"and":1,"tasted":1,"cri sp":1,"shared":1,"with":1,"her":2,"friend":2,"saying":1,"joy":1,"to":1,"enjoy":1,"natures":1,"gift":1,"replied":1,"perfect":1,"day":1,"for":1,"picnic":1};

[0180] The process of adding learning materials to the database includes:

[0181] (1) Generate a unique identifier for the material, material_id, using an auto-incrementing ID, UUID (Universally Unique Identifier), or a hash algorithm;

[0182] (2) Integrate the material content, vocabulary dictionary, grade, learning type and ID into a structured object (such as a dictionary or JSON format);

[0183] {"material_id":"MAT_READ_ENG_GR3_20241015_001",

[0184] "title":"An Apple Adventure",

[0185] "content":"What a delicious apple! Emma picked a bright red apple from the tree. She wondered, What kind of apple is this? It smelled sweet and tasted crisp. Emma shared the apple with her friend, saying, 'What a joy to enjoynature's gift!' Her friend replied, 'What a perfect day for an apple picnic!'",

[0186] "word_dict":{"what":4,"a":4,"delicious":1,"apple":4,"emma":2,"picked":1,"br ight":1,"red":1,"from":1,"the":3,"tree":1,"she":2,"wondered":1,"kind":1,"is":1,"this":1,"it":1,"smelled":1,"sweet":1," and":1,"tasted":1,"crisp":1,"shared":1,"with":1,"her":2,"friend":2,"saying":1,"joy":1,"to":1,"enjoy":1,"natures":1,"gif t":1,"replied":1,"perfect":1,"day":1,"for":1,"picnic":1},

[0187] "grade_level":"Grade 3",

[0188] "learning_type":"reading",

[0189] "keywords":["apple","nature","friendship"]}

[0190] (3) The structured objects are saved into the database for easy use later; the database can be MySQL, SQL Server, Oracle, etc. Any database that can achieve persistent storage is protected, and no specific limitation is made here.

[0191] The above data entry and structuring processes are just examples and can be modified according to actual needs.

[0192] It is worth noting that the feature vectors of all words and the word recommendation dictionary WordRecommendDict are also persistently stored in files or databases for easy use later.

[0193] S3. Dynamically update the word feature vector and recommendation value dictionary based on the student's performance in different learning types during the learning process; the learning types include word learning, speaking, dictation, listening, reading, and writing;

[0194] Based on students' performance in vocabulary learning, speaking, dictation, listening, reading, and writing, the feature vectors of each word are updated, and the recommended value dictionary is updated accordingly.

[0195] In this implementation, students learn English through a specific application on their tablets or mobile phones. The application provides feedback to understand students' mastery of vocabulary and updates the feature vectors and recommendation values ​​of the vocabulary dictionary.

[0196] It is understandable that the aforementioned specific applications can capture students' actions on the application page, such as recording voice or capturing students' writing actions in the application, such as writing start time and end time. This is existing technology and will not be elaborated here.

[0197] Different learning types update word feature vectors and the recommendation dictionary in different ways, including the following:

[0198] In the process of word learning, reading aloud and spelling, the inverse of the cumulative occurrence of the word in the corresponding feature vector, the inverse of the mastery level at the last learning time, and the timestamp of the last learning time are updated, and the word recommendation value is recalculated, thereby updating the word recommendation value dictionary WordRecommendDict.

[0199] The reciprocal of the cumulative occurrence count of a word is obtained by taking the reciprocal of the reciprocal of the cumulative occurrence count of a word in the word's feature vector. The cumulative occurrence count of a word is then incremented by 1 and the reciprocal is taken to obtain the latest reciprocal of the cumulative occurrence count of a word.

[0200] The timestamp of the last learning session is updated to the timestamp when the current word was learned.

[0201] The specific update process for the reciprocal of the mastery level from the last learning session is as follows:

[0202] Specifically, vocabulary learning often includes several steps such as listening, repeating, and spelling. We can calculate the student's current mastery of a word by comprehensively considering the accuracy of repeating (accuracy score Fl), the correctness of spelling (spelling accuracy score Sp), and the time spent spelling (time efficiency score Te).

[0203] Specifically, third-party speech recognition technologies are used to score the pronunciation of the words being read aloud, such as Tencent Cloud's Smart Listening Oral Evaluation (SOE), iFlytek's Spark Speech Big Model, and Unisound's AI Open Platform.

[0204] Smart Oral Evaluation (SOE) is a Chinese and English speech evaluation product launched by Tencent Cloud. It supports speech evaluation for all ages, from children to adults, and offers multiple modes such as word, sentence, paragraph, and free speech. It supports a comprehensive scoring mechanism that considers pronunciation accuracy, fluency, and completeness, with a similarity of over 95% to expert scores. It can be widely used in Chinese and English oral teaching.

[0205] Preferably, in this embodiment, the Smart Listening Oral Assessment is used to score the correct word pronunciation score Fl.

[0206] The Zhiling oral assessment can obtain a suggested score for the recording of an English word based on the recording of the word and the corresponding text. The suggested score is an integer from 1 to 100, which is used as the score for correct word pronunciation.

[0207] For example, the word we want to learn is "example". Based on the recording of the pronunciation and the accuracy score of "example" obtained through the Zhiling oral assessment, the pronunciation accuracy score is 96.

[0208] Furthermore, the spelling accuracy score Sp can be determined based on the proportion of correctly spelled letters in a word to the total number of letters in the word, with 100 representing completely correct spelling; the formula for calculating the spelling accuracy score Sp is shown in formula (3):

[0209] Sp=[(correctNum / totalNum)*100] (3)

[0210] Where Sp is the spelling accuracy score, correctNum is the number of correctly spelled letters, totalNum is the total number of letters in the word, and [] is the rounding symbol;

[0211] For example, we want to evaluate the spelling accuracy score Sp of the word "example".

[0212] Target word: example (7 letters), current spelling: example (note that the positions of 'l' and the second 'e' are reversed); compare each letter and count the number of correctly spelled letters. In "example", 'e', ​​'x', 'a', 'm', 'p' are correct, while the positions of 'e' and 'l' are reversed, so 'e' is considered correct (if we consider cases where position is not sensitive, but here we strictly calculate based on position). In fact, 'l' is incorrect, and the second 'e' is also incorrect. Therefore, the correct letters are 'e', ​​'x', 'a', 'm', 'p', a total of 5. Therefore, the spelling accuracy score Sp = [(5 / 7) * 100] = 71;

[0213] The time efficiency score Te is determined based on the time it takes to spell a word correctly, as shown in formula (4);

[0214] Te = Min[(T / t)*100, 100] (4)

[0215] Where Te is the time efficiency score, T = n * T1, T1 is the experience time to input an English letter on an electronic device, in ms, n represents the number of letters in a word, and t is the actual time the user spends to input a word, in ms.

[0216] The value of T1 is generally between 150-300ms, and in this embodiment, T1 is 250ms;

[0217] Taking the spelling of the word "example" as an example, "example" has 7 letters, so n = 7. T1 = 250ms, so T = 7 * 250 = 1750ms. Assuming the user took 2500ms, then Te = (1750 / 2500) * 100 = 70, so the score is 70. If the user took less time, such as 1500ms, then Te = (1750 / 1500) * 100 = 116, but according to the formula, Te takes the minimum value of 100, so the score is 100.

[0218] The current mastery level Mn of the word is calculated according to formula (5):

[0219] Mn=[Fl*a+Sp*b+Te*c] (5)

[0220] Where a is the weight of the reading accuracy score, b is the weight of the spelling accuracy score, and c is the weight of the time efficiency score; a+b+c=1;

[0221] Based on the principles of language learning, pronunciation and spelling are the core, while time efficiency is secondary. Therefore, the weight of accurate reading and spelling accuracy should be relatively large, while the weight of time efficiency can be secondary. Thus, in this embodiment, a = 0.4; b = 0.4; c = 0.2.

[0222] For example, if a student learns the word "example" and scores 96 for accurate pronunciation, 71 for spelling accuracy, and 70 for time efficiency, then the student's current mastery level of the word "example" is Mn = [96*0.4 + 71*0.4 + 70*0.2] = 80.

[0223] The formula for calculating the reciprocal v4' of the latest level of mastery is shown in formula (6):

[0224] v4'=R*v4+(1-R)*1 / Mn (6)

[0225] Where v4' is the reciprocal of the latest mastery level, v4 is the reciprocal of the last learning mastery level in the feature vector of the word, R is the Ebbinghaus forgetting curve, i.e.: R = e^(-t / S); where e is the base of the natural logarithm, t is the forgetting time in days, t = (ts - v5) / 86400; S is the relative memory strength, with a value of 1.84; ts is the current timestamp, and v5 is the timestamp of the last learning in the feature vector;

[0226] Taking the word "example" as an example, the reciprocal of the last learning mastery level of this word v4 = 0.1 (that is, the last learning mastery level was 10), the last learning timestamp is v5 = 1700000000, the current timestamp is ts = 1700432000, and the current mastery level Mn = 80;

[0227] First, calculate the forgetting time t = (1700432000 - 1700000000) / 86400 = 5. Then, R = e^(-5 / 1.84) ≈ 0.0655. Substitute this into the formula to get the reciprocal of the latest mastery level:

[0228] v4'=0.0655*0.1+(1-0.0655)*1 / 80≈0.00655+0.01168=0.01823;

[0229] This embodiment analyzes the above process:

[0230] Since the forgetting period is 5 days, the memory retention rate R≈6.55%, which shows an extremely low retention rate, indicating that most of the memory has decayed. The previous learning mastery level was 10, indicating that the previous learning effect was average and the word was basically not mastered. The current mastery level is 80, indicating a high level of mastery. The reciprocal of the current mastery level, 0.0125, has a significant impact on the final result. The latest mastery level is equal to 1 / v4'≈54.85, indicating that although the previous learning effect was average, the latest mastery level has improved to basic mastery through this learning.

[0231] Then replace the inverse of the latest mastery level with the inverse of the mastery level at the time of the last learning of the word feature vector;

[0232] Finally, the word recommendation value WR is recalculated based on the updated word feature vector using the formula for calculating word recommendation value, and then the word recommendation value dictionary WordRecommendDict is updated.

[0233] Oral language learning: Based on speech recognition technology, the pronunciation score of each word in the sentence is identified and read aloud. The current mastery level of each word is calculated based on the pronunciation score of each word. Then, the reciprocal of the latest mastery level of the word is calculated according to formula (6). The reciprocal of the cumulative occurrence of the word in the corresponding feature vector, the reciprocal of the mastery level at the last learning time, and the timestamp of the last learning time are updated. Then, the word recommendation value is recalculated, and the word recommendation value dictionary WordRecommendDict is updated.

[0234] The process of updating the cumulative number of times a word appears with the timestamp of the last learning session is the same as in the word learning process, and will not be repeated here.

[0235] Oral language learning is mainly assessed by evaluating students' learning effectiveness through recordings of them repeating sentences in English. Similarly, third-party speech recognition technology can be used to score the words in the recorded English sentences.

[0236] In addition to scoring individual words, the ZhiLing spoken English assessment can also score an entire English audio recording and its corresponding text. Based on the input English audio recording and its corresponding text, ZhiLing returns an overall score for the recording, as well as scores for each phoneme within the sentence (each phoneme carries its corresponding pronunciation letter information). From the words in the sentence and the letter information carried by each phoneme, the correspondence between each word and its corresponding phoneme can be determined. The average score of each phoneme is then used to obtain the word's pronunciation score.

[0237] The pronunciation score of the word is used as the current mastery level Mn of the word. Then, the reciprocal v4' of the latest mastery level of the word is calculated according to formula (6). The reciprocal of the latest mastery level is then used to replace the reciprocal of the mastery level of the word feature vector at the time of the last learning. Finally, the word recommendation value WR of the word is recalculated according to the updated word feature vector using the formula for calculating the word recommendation value, and then the word recommendation value dictionary WordRecommendDict is updated.

[0238] Dictation learning uses different methods to determine the current mastery level of words based on the spelling accuracy score. It calculates the reciprocal of the latest mastery level of a word based on the current mastery level, and then updates the reciprocal of the cumulative occurrence count of the word in the corresponding feature vector, the reciprocal of the mastery level at the time of the last learning, and the timestamp of the last learning. It then recalculates the word recommendation value and updates the word recommendation value dictionary WordRecommendDict.

[0239] The formula for calculating the spelling accuracy score Sp is shown in formula (3):

[0240] Sp=[(correctNum / totalNum)*100] (3)

[0241] The spelling accuracy score Sp is determined by the proportion of correct letters in a word to the total number of letters in the word, with 100 representing completely correct spelling.

[0242] When the word is spelled completely correctly, the time efficiency score Te is calculated using formula (4) and used as the current mastery level Mn of the word, i.e., Mn = Te; formula (4) is shown below;

[0243] Te = Min[(T / t)*100, 100] (4)

[0244] Where Te is the time efficiency score, T = n * T1, T1 is the experience time to input an English letter on an electronic device, in ms, n represents the number of letters in a word, and t is the actual time the user spends to input a word, in ms.

[0245] When a word is not spelled completely correctly, the current mastery level Mn of the word is calculated using formula (7):

[0246] Mn=U1*Sp+(1-U1)*Te (7)

[0247] Wherein, U1 is an empirical value, a floating-point number between 0 and 1. Considering that spelling words correctly is a better indicator of word mastery than the time spent on spelling, U1 can be larger. Therefore, in this embodiment, U1 is 0.75.

[0248] The process of updating the cumulative number of times a word appears with the timestamp of the last learning session is the same as in the word learning process, and will not be repeated here.

[0249] The current mastery level Mn of the word is calculated, and the reciprocal v4' of the latest mastery level of the word is calculated according to formula (6). Then, the reciprocal of the latest mastery level replaces the reciprocal of the mastery level of the word feature vector at the time of the last learning. Finally, the word recommendation value WR of the word is recalculated according to the updated word feature vector using the formula for calculating the word recommendation value, and then the word recommendation value dictionary WordRecommendDict is updated.

[0250] In writing learning, when a word in the writing content is in the WordRecommendDict dictionary and is spelled correctly, the inverse of the mastery level of the word at the last learning time is calculated. Then, the inverse of the cumulative occurrence of the word in the corresponding feature vector, the inverse of the mastery level at the last learning time, and the timestamp of the last learning time are updated, and the word recommendation value is recalculated, thereby updating the WordRecommendDict dictionary.

[0251] When learning to write, specific applications will recognize the handwriting of the user and generate corresponding letters. For example, they can use OCR technology or a pre-trained handwriting recognition model to recognize the handwriting and obtain the closest letters. Then, they will generate words based on the closest letter combinations. This is existing technology and will not be elaborated here. In addition, the start and end times of spelling are recorded for each word written.

[0252] Since students' spelling of words during writing is arbitrary, there is no answer that can determine whether the spelling of a word is correct. Therefore, this example only considers correctly spelled words.

[0253] Once the writing is complete, obtain all the words based on the content, and then iterate through all the words:

[0254] If a word is in the WordRecommendDict dictionary, the computation time efficiency score Te is calculated according to the time spent inputting the word using formula (4), and this score is used as the current mastery level Mn of the word. The current mastery level Mn is calculated using formula (6) to calculate the reciprocal v4' of the latest mastery level of the word. The reciprocal of the latest mastery level is then used to replace the reciprocal of the mastery level of the word feature vector at the time of the last learning. Finally, the word recommendation value WR of the word is recalculated according to the updated word feature vector using the word recommendation value calculation formula, and the WordRecommendDict dictionary is updated accordingly.

[0255] The process of updating the cumulative number of times a word appears with the timestamp of the last learning session is the same as in the word learning process, and will not be repeated here.

[0256] If a word is not in the WordRecommendDict dictionary, it is either misspelled, not an English word, or beyond the scope of primary and secondary school learning. You can simply ignore it.

[0257] Listening and reading learning

[0258] Listening learning generally involves listening to a long audio material and then choosing the correct answer based on the questions. If the answer is correct, it is assumed that all the words in the audio material have been learned once in the WordRecommendDict dictionary. Then, the inverse of the cumulative occurrence count of the words in the feature vector corresponding to these words and the timestamp of the last learning are updated. The current mastery level Mn of the word is set to 100. Then, the inverse of the latest mastery level of the word is calculated according to formula (6) based on the current mastery level Mn of the word. Finally, the words in the audio material in the WordRecommendDict dictionary are updated. The timestamp of the last learning after the update is taken as the timestamp when the audio material was learned.

[0259] Specifically, if the listening answers are completely correct, iterate through all the words in the WordDict of the listening material. If the word iterates through is in the recommendation value dictionary WordRecommendDict, then update the reciprocal of the cumulative occurrence count of the word in the feature vector corresponding to the word and the timestamp of the last learning, and set the current mastery level Mn of the word to 100.

[0260] The reciprocal v4' of the latest mastery level of a word is calculated using formula (6) based on the current mastery level Mn of the word. Then, the reciprocal of the latest mastery level replaces the reciprocal of the mastery level of the word feature vector at the time of the last learning. Finally, the word recommendation value WR of the word is recalculated based on the updated word feature vector using the formula for calculating the word recommendation value, and then the word recommendation value dictionary WordRecommendDict is updated.

[0261] If a student doesn't achieve a perfect score in listening comprehension, it's difficult to determine which words or words they didn't understand. In this case, we iterate through the WordDict and WordRecommendDict of the listening material to find the n words with the highest recommendation value, treating them as words they didn't understand. We then select a small percentage (e.g., 5%) of these n words based on the WordDict.

[0262] It is understandable that there will be corresponding written materials for the listening materials.

[0263] Specifically, if the listening answers are not entirely correct, the system iterates through the WordRecommendDict dictionary of all words in the listening material's WordDict, identifies the n words with the highest recommendation values ​​as the ununderstood words, and updates the feature vector of the ununderstood words with the reciprocal of the cumulative occurrence count and the timestamp of the last learning. The value of n is determined based on the length of the WordDict of the listening material, and is an integer obtained by rounding up to 5% of the length of the WordDict. The value of n can also be adjusted according to specific needs.

[0264] Because of the listening and reading materials, we cannot accurately know which words students did not master because they did not get full marks. We estimated based on the words with the highest recommended scores, so there may be a large error. Therefore, when calculating the reciprocal of the final mastery level of these words, we need to introduce an adjustment factor R1 to calculate the reciprocal of the latest mastery level v4'; the calculation formula is shown in formula (8):

[0265] v4'=R1*R*v3+(1-R1)*(1-R)*1 / Mn (8)

[0266] Where v4' is the reciprocal of the latest mastery level, v4 is the reciprocal of the last learning mastery level in the word's feature vector, R is the Ebbinghaus forgetting curve, i.e.: R = e^(-t / S); where e is the base of the natural logarithm, t is the forgetting time in days, t = (ts - v5) / 86400; S is the relative memory strength, with a value of 1.84; ts is the current timestamp, taken as the timestamp when the listening material was completed; v5 is the timestamp of the last learning in the word's feature vector; Mn is the current mastery level of the word; the value of Mn is determined by multiplying the final listening score by 100 as a percentage of the total score.

[0267] For example, if the total score of the current listening material is 60 points and the final score is 48 points, then the latest mastery level Mn = 48 / 60*100 = 80;

[0268] R1 is the memory influence coefficient. Considering that the inverse of the previous learning level has a greater influence on listening learning and the inverse of the most recent learning level has a smaller influence, the memory influence coefficient R1 in this embodiment is taken as 0.85.

[0269] Then, the reciprocal v4' of the latest mastery level is replaced with the reciprocal of the mastery level at the time of the last learning in the word feature vector; finally, the word recommendation value WR of the word is recalculated based on the updated word feature vector using the formula for calculating the word recommendation value, and then the word recommendation value dictionary WordRecommendDict is updated.

[0270] The process of updating the cumulative number of times a word appears in listening practice with the timestamp of the last learning session is the same as in vocabulary learning, and will not be repeated here.

[0271] The processing method for reading learning is similar to that for listening learning. Reading learning generally involves reading a long passage of reading material and then selecting the correct answer based on the questions. If the answer is correct, it is assumed that all the words in the reading material in the WordRecommendDict have been learned once. Then, the inverse of the cumulative occurrence count of the words in the feature vector corresponding to these words and the timestamp of the last learning are updated, and the current mastery level Mn of the word is set to 100. Then, the inverse of the latest mastery level of the word is calculated according to formula (6) based on the current mastery level Mn of the word. Finally, the words in the reading material in the WordRecommendDict are updated. The timestamp of the last learning after the update is taken as the timestamp when the reading material was completed.

[0272] Specifically, if the listening answers are completely correct, iterate through all the words in the word dictionary WordDict of the reading material. If the word iterates through is in the recommendation value dictionary WordRecommendDict, then update the inverse of the cumulative occurrence count of the word in the feature vector corresponding to the word and the timestamp of the last learning, and set the current mastery level Mn of the word to 100.

[0273] The reciprocal v4' of the latest mastery level of a word is calculated using formula (6) based on the current mastery level Mn of the word. Then, the reciprocal of the latest mastery level replaces the reciprocal of the mastery level of the word feature vector at the time of the last learning. Finally, the word recommendation value WR of the word is recalculated based on the updated word feature vector using the formula for calculating the word recommendation value, and then the word recommendation value dictionary WordRecommendDict is updated.

[0274] If a student doesn't achieve a perfect score in reading comprehension, it's difficult to determine which words or words they didn't understand. In this case, we iterate through the WordRecommendDict dictionary of the reading material to find the n words with the highest recommendation value, treating them as words they didn't understand. The n number is a small percentage (e.g., 5%) selected based on the length of the WordDict dictionary.

[0275] Specifically, if the listening answers are not entirely correct, the word recommendation value dictionary WordRecommendDict is traversed based on all words in the WordDict of the reading material to find the n words with the highest recommendation values ​​as the words that were not understood. The feature vectors of the words that were not understood are then updated with the reciprocal of the cumulative occurrence count of the words and the timestamp of the last learning. The value of n is determined based on the dictionary length of the WordDict of the reading material, and n is taken as an integer obtained by rounding up 5% of the dictionary length of WordDict.

[0276] For students learning reading materials, we also cannot accurately know which words students did not master because they did not get full marks. We also estimate based on the words with the highest recommended value, so there may be a large error. Therefore, when calculating the reciprocal of the final mastery level of these words, we need to introduce the memory influence coefficient R1 as a moderating factor to calculate the reciprocal of the latest mastery level v4'; the calculation formula is shown in formula (8):

[0277] v4'=R1*R*v3+(1-R1)*(1-R)*1 / Mn (8)

[0278] Where v4' is the reciprocal of the latest mastery level, v4 is the reciprocal of the last learning mastery level in the word's feature vector, R is the Ebbinghaus forgetting curve, i.e., R = e^(-t / S); where e is the base of the natural logarithm, t is the forgetting time in days, t = (ts - v5) / 86400; S is the relative memory strength, with a value of 1.84; ts is the current timestamp, taken as the timestamp when the reading material was completed; v5 is the timestamp of the last learning in the word's feature vector; Mn is the current mastery level of the word; the value of Mn is determined by multiplying the final listening score by 100.

[0279] For example, if the total score of the current reading material is 60 points and the final score is 48 points, then the latest mastery level Mn = 48 / 60 * 100 = 80;

[0280] R1 is the memory influence coefficient. Considering that the inverse of the previous learning level has a greater influence on reading and learning, while the inverse of the most recent learning level has a smaller influence, the memory influence coefficient R1 in this embodiment is set to 0.85.

[0281] Then, the reciprocal v4' of the latest mastery level is replaced with the reciprocal of the mastery level at the time of the last learning in the word feature vector; finally, the word recommendation value WR of the word is recalculated based on the updated word feature vector using the formula for calculating the word recommendation value, and then the word recommendation value dictionary WordRecommendDict is updated.

[0282] The process of updating the cumulative number of times a word appears in reading and the timestamp of the last learning session is the same as in word learning, and will not be repeated here.

[0283] S4. Select alternative learning materials based on grade level and learning type. Calculate the recommendation value of the alternative learning materials using the WordDict dictionary and the WordRecommendDict dictionary. Push the learning materials with the highest recommendation value to the students.

[0284] Specifically, it includes:

[0285] S41. When recommending learning materials, first select learning materials suitable for the student's age group as alternative learning materials based on the student's grade information and learning type.

[0286] The learning materials are stored with grade and learning type when they are added to the database, so they can be retrieved very easily. This is existing technology and will not be elaborated on here.

[0287] S42. Based on all the words in the vocabulary dictionary of the candidate learning materials, traverse the word recommendation value dictionary, find the m words with the highest word recommendation values, and generate a material recommendation dictionary;

[0288] The format of the material recommendation dictionary is the same as that of the word dictionary WordDict. The m words with the highest recommendation values ​​are selected from the word dictionary WordDict and then placed into the material recommendation dictionary according to the cumulative frequency of the words in the learning materials from largest to smallest.

[0289] The value of m is determined based on the length of the WordDict dictionary of the alternative learning materials. m is taken as 10% of the length of the WordDict dictionary and then rounded up to the nearest integer. The value of m can also be changed according to the needs, such as taking 15% or 20% of the length of the WordDict dictionary. There is no limitation here.

[0290] S43. If the learning type is oral, dictation, listening, reading or writing, the total recommendation value of the learning materials is calculated according to the material recommendation dictionary and formula (9), and the alternative learning materials with the highest total recommendation value are pushed to the students.

[0291]

[0292] Where MLR is the total recommendation score for the learning materials, count i The recommended word in the dictionary is the cumulative number of times it appears in the learning materials, where i is the ranking of the word's occurrences; i = 1 indicates the highest ranking; WR i Recommend the word value for the i-th word in the dictionary for the material;

[0293] For example, by filtering all learning materials from the learning materials database that are for the third grade and whose learning type is "reading", a set of candidate materials is obtained. There are two candidate learning materials, namely Material A and Material B.

[0294] Material A: WordDict = {"apple":5,"what":2,"hello":3...}, a total of 20 key-value pairs;

[0295] Material B: WordDict = {"apple":3,"what":4,"test":2...}, a total of 20 key-value pairs;

[0296] WordRecommendDict={"hello":0.485,"apple":0.378,"what":0.395};

[0297] m = 20 * 10% = 2 for both materials;

[0298] Regarding material A:

[0299] WordDict sorts words according to their WordRecommendDict values, resulting in:

[0300] "hello" (WR=0.485) (Recommended value #1), "what" (WR=0.395) (Recommended value #2), "apple" (WR=0.378) (Recommended value #3), ..., then take the first two words to get the recommended dictionary A = {"hello":3,"what":2};

[0301] For material B:

[0302] WordDict sorts words according to their WordRecommendDict values, resulting in:

[0303] "what" (WR=0.395) (Recommended value 1) > "apple" (WR=0.378) (Recommended value 2) > "test" (0) (Recommended value 3), ..., then take the first 2 words to get the material recommended dictionary B={"what":4,"apple":3};

[0304] The total recommended learning score (MLR_A) for material A is calculated as follows: MLR_A = 3 / 1 * 0.485 + 2 / 2 * 0.395 = 1.85.

[0305] The total recommended learning score (MLR_B) for material B is calculated as follows: MLR_B = 4 / 1 * 0.395 + 3 / 2 * 0.378 = 2.336.

[0306] It can be seen that the overall recommendation value of material B is greater than that of material A, therefore material B is recommended to students for learning.

[0307] The dynamic weighting formula for the total recommendation value of learning materials in this embodiment balances word recommendation values ​​with actual usage frequency. First, it sorts the words by their WordRecommendDict recommendation values ​​and selects the top m words to prioritize high-recommendation vocabulary. Then, it uses dynamic weighting... By balancing the recommended word value with the actual frequency of its appearance in the materials, the system can accurately match students' needs and prioritize pushing materials with high recommendation values ​​and high frequency of appearance. This recommendation method combines the difficulty, importance, mastery level, and forgetting patterns of words to achieve more optimized personalized learning recommendations.

[0308] S44. If the learning type is word learning, the total recommendation value of the learning materials is calculated according to the material recommendation dictionary and formula (9). The words in the word dictionary of the candidate learning materials with the highest total recommendation value are sorted from high to low according to their recommendation values ​​in the word recommendation value dictionary and then pushed to the students.

[0309] Furthermore, after sorting the words in the WordDict of the candidate learning materials with the highest total recommendation value from highest to lowest in WordRecommendDict, users can select a preset number of words to push to the system. The preset number can be set to 5 by default or changed according to user needs.

[0310] Steps S3 and S4 form a closed loop. Each time, students receive learning materials containing their weakest words and study them. The learning results are fed back to the feature vectors and recommendation values ​​of each word. This information is then used to determine the content of the next learning material. This process is repeated, which greatly improves the efficiency of students' English learning.

[0311] This embodiment addresses the problems of low efficiency, lack of personalized dynamic recommendation mechanisms, and difficulty in adapting to multimodal learning scenarios and individual memory decay patterns in traditional vocabulary learning. It identifies words that need to be strengthened from daily vocabulary learning, speaking, dictation, listening, reading, and writing, and achieves a closed-loop optimization of long-term learning effects through dynamic feature engineering and multi-scenario recommendation, effectively improving the efficiency of vocabulary learning.

[0312] Furthermore, traditional vocabulary learning uses a fixed order or random repetition, failing to consider the dynamic changes in word difficulty and student mastery, leading to inefficient repetition or omission of key words. This embodiment quantifies word learning priorities through feature vectors (such as syllable count, importance, and reciprocal of cumulative occurrences), and adjusts recommendation values ​​based on student dynamic performance to achieve precise learning. Through a dynamic recommendation value dictionary, words with low student mastery, high importance, and suitability for the current learning type are prioritized, reducing ineffective repetition and shortening the time to achieve the goal. Multi-dimensional feature vectors (such as timestamp decay memory curves and reciprocal of cumulative occurrences to avoid "high-frequency word monopoly") replace subjective experience-based ranking, making the learning path more in line with cognitive patterns.

[0313] Furthermore, traditional methods lack a real-time feedback mechanism for students' learning behavior and cannot dynamically adjust learning strategies. This embodiment dynamically updates word feature vectors in step S3 (e.g., adjusting the "reciprocal of the previous mastery level" based on the dictation error rate) to achieve continuous optimization of the learning path. Combined with performance data from scenarios such as speaking and dictation, the word recommendation strategy is optimized in reverse (e.g., increasing the audio training weight for words with listening errors), forming a "learning-evaluation-optimization" closed loop.

[0314] Furthermore, by using the timestamps in the feature vectors and the reciprocal of the mastery level, the word memory decay curve can be tracked, providing data support for subsequent review plans.

[0315] Example 2

[0316] refer to Figure 2 This embodiment provides a device for improving the vocabulary learning efficiency of primary and secondary school students. Based on the above-mentioned method for improving the vocabulary learning efficiency of primary and secondary school students, it includes the following units:

[0317] An initialization unit is used to define the feature vector of a word and the word list of the learning material; the feature vector includes the number of syllables of the word, the word importance, the reciprocal of the cumulative number of times the word appears, the reciprocal of the mastery level of the last learning, and the timestamp of the last learning.

[0318] The construction unit is used to build feature vectors for all words in the English learning materials according to the syllabus requirements and textbook content, then calculate word recommendation values ​​based on the feature vectors, initialize the word recommendation value dictionary, and input the learning materials into the database; the learning materials include the corresponding word dictionary;

[0319] The update unit is used to dynamically update the word feature vector and recommendation value dictionary based on the student's performance in different learning types during the learning process; the learning types include word learning, oral communication, dictation, listening comprehension, reading, and writing.

[0320] The recommendation unit is used to filter out alternative learning materials based on grade level and learning type. It calculates the recommendation value of the alternative learning materials using a vocabulary dictionary and a word recommendation value dictionary, and pushes the learning materials with the highest recommendation value to the students.

[0321] Example 3

[0322] refer to Figure 3 , Figure 3 This is a schematic diagram of the structure of a device for improving vocabulary learning efficiency for primary and secondary school students according to this embodiment. The device 20 for improving vocabulary learning efficiency for primary and secondary school students in this embodiment includes a processor 21, a memory 22, and a computer program stored in the memory 22 and executable on the processor 21. When the processor 21 executes the computer program, it implements the steps in the above method embodiments. Alternatively, when the processor 21 executes the computer program, it implements the functions of each module / unit in the above device embodiments.

[0323] For example, the computer program can be divided into one or more modules / units, which are stored in the memory 22 and executed by the processor 21 to complete the present invention. The one or more modules / units can be a series of computer program instruction segments capable of performing specific functions, which describe the execution process of the computer program in the device 20 for improving the vocabulary learning efficiency of primary and secondary school students. For example, the computer program can be divided into the modules shown in Embodiment 2. The specific functions of each module are described in the working process of the device described in the above embodiments, and will not be repeated here.

[0324] The device 20 for improving vocabulary learning efficiency for primary and secondary school students may include, but is not limited to, a processor 21 and a memory 22. Those skilled in the art will understand that the schematic diagram is merely an example of the device 20 for improving vocabulary learning efficiency for primary and secondary school students and does not constitute a limitation on the device 20. It may include more or fewer components than illustrated, or combine certain components, or use different components. For example, the device 20 for improving vocabulary learning efficiency for primary and secondary school students may also include input / output devices, network access devices, buses, etc.

[0325] The processor 21 can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor. The processor 21 is the control center of the device 20 for improving the vocabulary learning efficiency of primary and secondary school students, connecting all parts of the device 20 through various interfaces and lines.

[0326] The memory 22 can be used to store the computer programs and / or modules. The processor 21 implements various functions of the device 20 for improving the vocabulary learning efficiency of primary and secondary school students by running or executing the computer programs and / or modules stored in the memory 22 and calling the data stored in the memory 22. The memory 22 may mainly include a program storage area and a data storage area. The program storage area may store the operating system, at least one application program required for a function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created according to the use of the mobile phone (such as audio data, phone book, etc.). In addition, the memory 22 may include high-speed random access memory, and may also include non-volatile memory, such as hard disk, memory, plug-in hard disk, smart media card (SMC), secure digital card (SD) card, flash card, at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0327] The integrated modules / units of the device 20 for improving vocabulary learning efficiency for primary and secondary school students, if implemented as software functional units and sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the above embodiments of the present invention can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by the processor 21, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc.

[0328] It should be noted that the device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Furthermore, in the accompanying drawings of the device embodiments provided by this invention, the connection relationships between modules indicate that they have communication connections, which can be specifically implemented as one or more communication buses or signal lines. Those skilled in the art can understand and implement this without any creative effort.

[0329] In this embodiment, the device can be a mobile phone or a tablet.

[0330] The parts of this invention not described in detail are prior art. It will be apparent to those skilled in the art that this invention is not limited to the details of the above exemplary embodiments, and that the invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the invention. Therefore, the embodiments should be regarded as exemplary and non-limiting in all respects, and are intended to encompass all changes falling within the meaning and scope of equivalents within this invention.

Claims

1. A method for improving vocabulary learning efficiency for primary and secondary school students, characterized in that, Specifically, the following steps are included: S1. Define the feature vector of a word and the word list of the learning material; the feature vector includes the number of syllables of the word, the word importance, the reciprocal of the cumulative number of times the word appears, the reciprocal of the mastery level of the last learning, and the timestamp of the last learning. S2. Based on the syllabus requirements and textbook content, establish feature vectors for all words in the English learning materials, then calculate word recommendation values ​​based on the feature vectors, initialize the word recommendation value dictionary, and input the learning materials into the database; the learning materials include the corresponding word dictionary; S3. Dynamically update the word feature vector and recommendation value dictionary based on the student's performance in different learning types during the learning process; the learning types include word learning, speaking, dictation, listening, reading, and writing; S4. Select alternative learning materials based on grade level and learning type. Calculate the recommendation value of the alternative learning materials using a vocabulary dictionary and a word recommendation value dictionary. Push the learning materials with the highest recommendation value to the students.

2. The method according to claim 1, characterized in that, In step S2, the feature vector is used to calculate the word recommendation value WR using formula (1), which is shown below: WR=W1*v1+W2*v2+W3*v4+W4*v4+W5*v5' (1) Wherein, v1 is the number of syllables in the word, and W1 is the weight of the number of syllables in the word; v2 is the importance of the word, and W2 is the weight of the importance of the word; v3 is the reciprocal of the cumulative number of occurrences of the word, and W3 is the weight of the reciprocal of the cumulative number of occurrences of the word; v4 is the reciprocal of the mastery level at the time of the last learning, and W4 is the weight of the reciprocal of the mastery level at the time of the last learning; v5' is the forgetting coefficient obtained by converting the timestamp v5 of the last learning, and W5 is the weight of the forgetting coefficient.

3. The method according to claim 2, characterized in that, The formula for calculating the forgetting coefficient is shown in formula (2): Fc=1-R (2) Where Fc is the forgetting coefficient; R is the Ebbinghaus forgetting curve, i.e.: R=e^(-t / S), where e is the base of the natural logarithm, t is the forgetting time in days, t=(ts-v5) / 86400; S is the relative memory strength; ts is the current timestamp, and v5 is the timestamp of the last learning in the feature vector.

4. The method according to claim 1, characterized in that, Step S3 specifically includes: In vocabulary learning, during the process of reading aloud and spelling, the system updates the corresponding feature vector by the reciprocal of the cumulative occurrence count of the word, the reciprocal of the mastery level at the time of the last learning session, and the timestamp of the last learning session. It then recalculates the word recommendation value and updates the word recommendation dictionary accordingly. Specifically: The reciprocal of the cumulative occurrence count of a word is obtained by taking the reciprocal of the reciprocal of the cumulative occurrence count of a word in the word's feature vector. The cumulative occurrence count of a word is then incremented by 1 and the reciprocal is taken to obtain the latest reciprocal of the cumulative occurrence count of a word. The timestamp of the last learning session is updated to the timestamp when the current word was learned. Based on the word pronunciation accuracy score Fl, spelling accuracy score Sp, and time efficiency score Te, the reciprocal v4' of the latest mastery level of the word is calculated using formulas (5)-(6). Then, the reciprocal of the mastery level of the word feature vector at the time of the last learning is replaced. Finally, the word recommendation value WR of the word is recalculated based on the updated word feature vector using the formula for calculating the word recommendation value, thereby updating the word recommendation value dictionary. Formulas (5)-(6) are shown below: Mn=[Fl*a+Sp*b+Te*c] (5) Where Mn represents the current mastery level of the word, a represents the weight of the reading accuracy score, b represents the weight of the spelling accuracy score, and c represents the weight of the time efficiency score; a+b+c=1; The formula for calculating the reciprocal v4' of the latest level of mastery is shown in formula (6): v4'=R*v4+(1-R)*1 / Mn (6) Where v4' is the reciprocal of the latest mastery level, v4 is the reciprocal of the last learning mastery level in the feature vector of the word, R is the Ebbinghaus forgetting curve, i.e.: R = e^(-t / S); where e is the base of the natural logarithm, t is the forgetting time in days, t = (ts-v5) / 86400; S is the relative memory strength, with a value of 1.84; ts is the current timestamp, and v5 is the timestamp of the last learning in the feature vector.

5. The method according to claim 4, characterized in that: The accuracy score of word repetition is obtained by scoring the pronunciation of the words being read aloud using third-party speech recognition technology; The spelling accuracy score Sp is determined according to formula (3): Sp=[(correctNum / totalNum)*100] (3) Where, correctNum is the number of letters spelled correctly, totalNum is the total number of letters in the word, and [] is the rounding symbol; the time efficiency score Te is determined based on the time to spell a correct word, as shown in formula (4); Te = Min[(T / t)*100, 100] (4) Where Te is the time efficiency score, T = n * T1, T1 is the experience time to input an English letter on an electronic device, in ms, n represents the number of letters in a word, and t is the actual time the user spends to input a word, in ms.

6. The method according to claim 4, characterized in that, Step S3 specifically also includes: In oral language learning, the pronunciation score of each word in the sentence is identified by speech recognition technology, and the current mastery level of each word is calculated based on the pronunciation score of each word. Then, the reciprocal of the latest mastery level of the word is calculated according to formula (6), and the reciprocal of the cumulative occurrence of the word in the corresponding feature vector, the reciprocal of the mastery level at the last learning time, and the timestamp of the last learning time are updated. Then, the word recommendation value is recalculated, and the word recommendation value dictionary is updated.

7. The method according to claim 5, characterized in that, Step S3 specifically also includes: Dictation learning uses different methods to determine the current mastery of words based on the spelling accuracy score. Based on the current mastery of words, the reciprocal of the latest mastery of words is calculated using formula (6). Then, the reciprocal of the cumulative occurrence of words in the corresponding feature vector, the reciprocal of the mastery level at the last learning time, and the timestamp of the last learning time are updated. The word recommendation value is recalculated, and the word recommendation value dictionary is updated. When the word is spelled completely correctly, the time efficiency score Te is calculated using formula (4) and used as the current mastery level Mn of the word, i.e., Mn = Te; When a word is not spelled completely correctly, the current mastery level Mn of the word is calculated using formula (7): Mn=U1*Sp+(1-U1)*Te (7) Where U1 is an empirical value, a floating-point number between 0 and 1.

8. The method according to claim 1, characterized in that, Step S3 specifically also includes: During listening / reading, if the listening / reading answers are completely correct, iterate through all the words in the vocabulary dictionary of the listening / reading material. If the word iterates through is in the recommendation value dictionary, update the reciprocal of the cumulative occurrence count of the word in the feature vector corresponding to the word and the timestamp of the last learning, and set the current mastery level Mn of the word to 100; then calculate the reciprocal v4' of the latest mastery level of the word according to formula (6) based on the current mastery level Mn of the word. If the listening / reading answers are not completely correct, the word recommendation value dictionary is traversed according to all words in the vocabulary dictionary of the listening / reading material, and the n words with the highest word recommendation value are identified as the words that were not understood. The reciprocal of the cumulative occurrence count of the word and the timestamp of the last learning are updated in the feature vector corresponding to the words that were not understood. The value of n is determined according to the length of the vocabulary dictionary of the listening material. The reciprocal of the latest mastery level v4' is calculated according to formula (8): v4'=R1*R*v3+(1-R1)*(1-R)*1 / Mn (8) Where v4' is the reciprocal of the latest mastery level, v4 is the reciprocal of the last learning mastery level in the feature vector of the word, R is the Ebbinghaus forgetting curve retention rate, Mn is the current mastery level of the word; the value of Mn is determined by multiplying the final score of the listening test by 100; R1 is the memory influence coefficient. Then, the reciprocal v4' of the latest mastery level is replaced with the reciprocal of the mastery level at the time of the last learning in the word feature vector; finally, the word recommendation value is recalculated based on the updated word feature vector using the formula for calculating the word recommendation value, thereby updating the word recommendation value dictionary.

9. The method according to claim 1, characterized in that, Step S4 specifically includes the following steps: S41. When recommending learning materials, first select learning materials suitable for the student's age group as alternative learning materials based on the student's grade information and learning type. S42. Based on all the words in the vocabulary dictionary of the candidate learning materials, traverse the word recommendation value dictionary, find the m words with the highest word recommendation values, and generate a material recommendation dictionary; S43. If the learning type is oral, dictation, listening, reading or writing, the total recommendation value of the learning materials is calculated according to the material recommendation dictionary and formula (9), and the alternative learning materials with the highest total recommendation value are pushed to the students. Where MLR is the total recommendation score for learning materials, count i The recommended word in the dictionary is the cumulative number of times it appears in the learning materials, where i is the ranking of the word's occurrences; i = 1 indicates the highest ranking; WR i Recommend the word value of the i-th word in the dictionary for the material.

10. A device for improving vocabulary learning efficiency for primary and secondary school students, characterized in that, Includes the following units: An initialization unit is used to define the feature vector of a word and the word list of the learning material; the feature vector includes the number of syllables of the word, the word importance, the reciprocal of the cumulative number of times the word appears, the reciprocal of the mastery level of the last learning, and the timestamp of the last learning. The construction unit is used to build feature vectors for all words in the English learning materials according to the syllabus requirements and textbook content, then calculate word recommendation values ​​based on the feature vectors, initialize the word recommendation value dictionary, and input the learning materials into the database; the learning materials include the corresponding word dictionary; The update unit is used to dynamically update the word feature vector and recommendation value dictionary based on the student's performance in different learning types during the learning process; the learning types include word learning, oral communication, dictation, listening comprehension, reading, and writing. The recommendation unit is used to filter candidate learning materials based on grade information and learning type. It calculates the recommendation value of the candidate learning materials by using a vocabulary dictionary and a word recommendation value dictionary, and pushes the learning materials with the highest recommendation value to students.

Images