Word recommendation method and device

Abstract

The application provides a word recommendation method and device, the word recommendation method comprises: obtaining a plurality of words to be recommended, the occurrence frequency of each word and the memory amount of each word of a user; determining the proficiency of the user to each word based on a prediction model; comprehensively evaluating the plurality of words based on the occurrence frequency, the proficiency and the memory amount of the word, and determining target recommended words according to the comprehensive evaluation result. The method can obtain the amount of words required to be recited by the user, improve the accuracy of word recommendation, and further improve the efficiency of word recitation of the user.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, and in particular to a word recommendation method and device. BACKGROUND

[0002] Word memorization is an indispensable part of English learning, and scientifically and effectively memorizing words can improve the efficiency of word memorization of users.

[0003] In related technologies, when users use word software or electronic dictionaries to recite words, the word library provided by the software itself is too extensive, and the amount of words required by the user cannot be accurately provided, resulting in that the user will spend a lot of time on simple and repeatedly memorized words when memorizing words, and the efficiency of word memorization is low. SUMMARY

[0004] The present application provides a word recommendation method and device to solve the defect that the prior art cannot accurately provide the amount of words required by the user, resulting in that the user will spend a lot of time on simple and repeatedly memorized words when reciting words, and improve the efficiency of word memorization of users.

[0005] The present application provides a word recommendation method, comprising:

[0006] Obtaining a plurality of words to be recommended, the occurrence frequency of each word, and the memory amount of each word by the user;

[0007] Determining the proficiency of each word by the user based on a prediction model, wherein the prediction model takes the feedback results of other users using a plurality of word exercises as training samples, and the type of the feedback results is obtained as a training label;

[0008] Comprehensively evaluating the plurality of words based on the occurrence frequency, the proficiency, and the word memory amount, and determining a target recommended word according to the comprehensive evaluation result, wherein the target recommended word is one of the plurality of words.

[0009] According to the word recommendation method provided by the present application, the comprehensive evaluation of the plurality of words based on the occurrence frequency, the proficiency, and the word memory amount, and the determination of the target recommended word according to the comprehensive evaluation result, comprises:

[0010] Determining the frequency score of each word based on the mapping relationship between the occurrence frequency and a preset frequency score weight;

[0011] Determining the proficiency score of each word by the user based on the mapping relationship between the proficiency and a preset proficiency score weight;

[0012] Determining the memory amount of each word by the user based on the mapping relationship between the memory amount and a preset memory amount score weight;

[0013] determine a comprehensive score of each word based on the frequency score of each word, the proficiency score of each word of the user and the memory amount score of each word of the user;

[0014] determine the word with the highest comprehensive score as the target recommended word based on the comprehensive scores of the words.

[0015] According to the word recommendation method provided by the application, the obtained multiple words to be recommended, the frequency of each word and the memory amount of each word of the user further comprise:

[0016] Based on the past years' examination true questions, the multiple words and the frequency of each word are obtained through word segmentation processing and statistics.

[0017] According to the word recommendation method provided by the application, the prediction model is obtained through the following steps:

[0018] Obtain the feedback results of multiple word exercises and the types of the multiple feedback results, and the types of the feedback results include correct use of words and incorrect use of words;

[0019] The word exercise includes a stem and multiple options, wherein the stem includes a word or a Chinese interpretation of a word, when the stem is a word, the multiple options are respectively Chinese interpretations associated with the word, when the stem is a Chinese interpretation of a word, the multiple options are respectively words associated with the Chinese interpretation, and the association between the multiple options is determined based on cosine similarity;

[0020] The feedback results of the multiple word exercises are one-hot encoded to obtain sample data;

[0021] The LSTM network is trained with the sample data as the training sample and the types of the feedback results as the training label to obtain the prediction model.

[0022] According to the word recommendation method provided by the application, the obtained multiple words to be recommended, the frequency of each word and the memory amount of each word of the user further comprise:

[0023] Based on the Ebbinghaus forgetting curve, the memory amount of each word of the user is determined.

[0024] According to the word recommendation method provided by the application, the comprehensive score of each word is determined based on the frequency score of each word, the proficiency score of each word of the user and the memory amount score of each word of the user, comprising:

[0025] The comprehensive score of each word is determined by applying the following formula:

[0026] R u = |I(u)| -0.5 x v∈I(u) x v ,

[0027] wherein R u is a comprehensive score, I(u) is implicit feedback of all words of user u, x v is an implicit feedback feature vector determined based on the frequency score, the proficiency score and the memory amount score, v is a word, and u is a user.

[0028] The application further provides a word recommendation device, comprising:

[0029] a word acquisition module, configured to acquire a plurality of words to be recommended, an occurrence frequency of each word and a memory amount of each word by a user;

[0030] a prediction module, configured to determine a proficiency of each word by the user based on a prediction model, wherein the prediction model takes feedback results of a plurality of words by other users as training samples and takes types of the feedback results as training labels;

[0031] a word recommendation module, configured to comprehensively evaluate the plurality of words based on the occurrence frequency, the proficiency and the memory amount of the words, and determine a target recommended word according to a comprehensive evaluation result, wherein the target recommended word is one of the plurality of words.

[0032] The application further provides an electronic device, comprising a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements the word recommendation method according to any one of the above when executing the program.

[0033] The application further provides a non-transitory computer readable storage medium, which stores a computer program, wherein the computer program is executable on a processor to implement the word recommendation method according to any one of the above.

[0034] The application further provides a computer program product, comprising a computer program, wherein the computer program is executable on a processor to implement the word recommendation method according to any one of the above.

[0035] The word recommendation method and device provided by the application can predict the proficiency of each word by the user through a prediction model, comprehensively evaluate each word based on the occurrence frequency, the proficiency and the memory amount of the word, and finally determine a target recommended word according to a comprehensive evaluation result, so that the amount of words to be recited by the user can be obtained, the accuracy of word recommendation is improved, and the efficiency of word recitation by the user is improved. BRIEF DESCRIPTION OF DRAWINGS

[0036] In order to more clearly illustrate the technical solutions of the present application or the prior art, the following will briefly introduce the drawings required by the embodiments or prior art description. Obviously, the drawings described below are only some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor.

[0037] Figure 1 is a flowchart of the word recommendation method provided by the present application;

[0038] Figure 2 is a correlation diagram of the "hidden" state, input vector sequence and output vector sequence based on the LSTM model provided by the present application;

[0039] Figure 3 is a structural diagram of the word recommendation device provided by the present application;

[0040] Figure 4 is a physical structure diagram of an electronic device provided by the present application. DETAILED DESCRIPTION

[0041] In order to make the objects, technical solutions and advantages of the present application clearer, the technical solutions of the present application will be described clearly and completely below in combination with the drawings in the present application. Obviously, the described embodiments are only some embodiments of the present application, not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0042] The word recommendation method and device of the present application will be described below in combination with Figures 1-3

[0043] Figure 1 is a flowchart of the word recommendation method provided by the present application, and the present application provides a word recommendation method, comprising the following steps:

[0044] Step 110, obtaining a plurality of words to be recommended, the frequency of occurrence of each word and the memory amount of each word by the user.

[0045] In this step, the plurality of words to be recommended refers to the words recommended by the terminal to the user for memorizing. The plurality of words can include English words that have been recited by the user, or English words that have not been recited by the user.

[0046] In this step, the frequency of occurrence of the word can be the number of times the word appears in the past years' examination true questions. For example, the terminal collects English past years' examination questions, performs word segmentation processing on the question content to obtain a word collection library, and counts the number of times the word appears in the past years' true questions. ​

[0047] In this step, the memory amount of each word of the user represents the degree of memory of each word of the user. Generally, after the user recites the word, the memory amount of the user to the word is large in a short period of time, for example, the memory amount of the user to the word remains above 0.7 within 1 minute to 10 minutes after the user finishes reciting the word; when the user does not repeatedly recite the word, the memory amount of the user to the word decreases with the increase of time, that is, the user will gradually forget the recited word.

[0048] In this embodiment, the frequency threshold can be set to divide the word frequency category into high-frequency words, medium-frequency words and low-frequency words, and the occurrence frequency of each word is compared with the frequency threshold to determine the corresponding frequency category of each word.

[0049] In this embodiment, the frequency threshold can be set according to user needs.

[0050] For example, the word occurrence frequency is less than 5 times, which belongs to low-frequency words, the occurrence frequency is in the range of 5 to 10, which belongs to medium-frequency words, and the occurrence frequency is above 10, which belongs to high-frequency words.

[0051] In some embodiments, the memory amount of each word of the user can be determined by using the Ebbinghaus forgetting curve.

[0052] Step 120, based on the prediction model, determine the proficiency of the user to each word, the prediction model uses the feedback results of other users using a plurality of word exercises as training samples, and the type of feedback results as training labels.

[0053] In this step, the prediction model can be trained by using the sample to the neural network.

[0054] In this embodiment, the neural network includes LSTM (Long Short-Term Memory), DNN (Deep Neural Networks) and the like.

[0055] In this step, the feedback result includes the result of the other user using the word exercise to answer the question, 0 represents the wrong answer, and 1 represents the correct answer. In this embodiment, the training label can be represented by numbers, letters or other symbols, for example, the training label can be represented by 0 and 1, 0 represents the first feedback result is wrong, and 1 represents the first feedback result is correct.

[0056] Step 130, based on the occurrence frequency, proficiency and word memory amount, the plurality of words are comprehensively evaluated, and the target recommended word is determined according to the comprehensive evaluation result, the target recommended word is one of the plurality of words.

[0057] In this step, the frequency of occurrence of the word, the proficiency of the user to the word and the memory amount of the user to the word can be used as three evaluation indexes to measure the importance of the word.

[0058] In this embodiment, the word can be scored according to the frequency of occurrence of the word, for example, the word A belongs to a high-frequency word, which indicates that the word A is applied more in the test questions, so the word A should be focused on, and the frequency score of the word A is counted as 3 points, if the word A is a medium-frequency word, the score is counted as 2, if the word A is a low-frequency word, the score is counted as 1, and so on.

[0059] In this embodiment, the predicted proficiency can also be scored, the higher the proficiency, the greater the probability that the user can master the word A, and a lower score can be set, the lower the proficiency, the smaller the probability that the user can master the word A, and the user needs to recite, and a higher score can be set.

[0060] In this embodiment, the word memory amount can also be scored, the smaller the word memory amount, the greater the degree of forgetting of the user to the word, which needs to be reviewed and consolidated in time, and a higher score can be set, the greater the word memory amount, the clearer the memory of the user to the word, and a lower score can be set.

[0061] After scoring each word from different dimensions such as frequency of occurrence, proficiency and word memory amount, the word with the highest score is a word with high frequency of occurrence, low proficiency of the user and low memory amount of the user to the word, which is also a word that needs to be recommended to the user in priority.

[0062] The word recommendation method provided by the embodiment of the application can predict the proficiency of the user to each word through a prediction model, comprehensively evaluate each word through the frequency of occurrence of the word, the proficiency and the word memory amount, and finally determine the target recommended word according to the comprehensive evaluation result, so that the amount of words to be recited by the user can be obtained, the accuracy of word recommendation is improved, and the efficiency of word recitation of the user is improved.

[0063] In some embodiments, the plurality of words are comprehensively evaluated based on the frequency of occurrence, the proficiency and the word memory amount, and the target recommended word is determined according to the comprehensive evaluation result, including: determining the frequency score of each word based on the mapping relationship between the frequency of occurrence and the preset frequency score weight; determining the proficiency score of the user to each word based on the mapping relationship between the proficiency and the preset proficiency score weight; determining the memory amount of the user to each word based on the mapping relationship between the memory amount and the preset memory amount score weight; determining the comprehensive score of each word based on the frequency score of each word, the proficiency score of the user to each word and the memory amount score of the user to each word; and determining the word with the highest comprehensive score as the target recommended word based on the comprehensive score of each word.

[0064] In this embodiment, the pre-frequency score weight can be set according to user needs, for example, the pre-frequency score weight is set as follows:

[0065]

[0066] wherein, P v is the frequency score of the word v.

[0067] In this embodiment, the way of determining whether the word to be recommended belongs to high frequency, medium frequency or low frequency is determined by the occurrence frequency of the word to be recommended and the above-mentioned frequency threshold, and this embodiment will not be described again; when the word A is a high-frequency word, the frequency score of the word A is 3.

[0068] In this embodiment, the proficiency score weight can be set according to user needs, for example, the proficiency score weight is set as follows:

[0069]

[0070] wherein, d uv is the proficiency of the user u to the word v, and D uv is the proficiency score of the user u to the word v.

[0071] In this embodiment, d uv is determined according to the prediction model; when the proficiency of the word A is 0.3, the corresponding proficiency score is 3.

[0072] In this embodiment, the memory amount score weight can be set according to user needs, for example, the pre-frequency score weight is set as follows:

[0073]

[0074] wherein, t uv is the memory time of the user u to the word v, and A uv is the memory amount score of the user u to the word v; the memory amount of the user to the word can be determined according to the memory time, for example, the memory amount of the user to the word at different memory time intervals can be determined according to the Ebbinghaus forgetting curve; when the memory time of the word A is 1.5 days, the corresponding memory amount score is 2.

[0075] In this embodiment, the comprehensive score corresponding to each word can be obtained by comprehensively calculating the frequency score, the proficiency score and the memory amount score of each word of the user.

[0076] In this embodiment, the comprehensive score can be the cumulative contribution degree of the occurrence frequency, the proficiency and the memory amount of each word, and then the comprehensive score of each word is obtained by normalization processing.

[0077] In some embodiments, the occurrence frequency, the proficiency and the memory amount of each word can also be weighted and summed with the corresponding frequency score, proficiency score and memory amount score respectively to obtain a comprehensive score of each word.

[0078] The word recommendation method provided by the embodiment of the application realizes the evaluation of the importance of the to-be-recommended words from the occurrence frequency, the proficiency and the memory amount of the words, and improves the accuracy of the recommended words for the user.

[0079] In some embodiments, obtaining the to-be-recommended words, the occurrence frequency of each word and the memory amount of each word by the user further comprises: obtaining the to-be-recommended words and the occurrence frequency of each word by word segmentation processing and statistics based on the past years' examination true questions.

[0080] In this embodiment, the occurrence frequency of the words can be obtained from the question bank of the past years' examination true questions, because the amount of words examined in the past years' examination true questions has important reference value.

[0081] In this embodiment, the number and type of words contained in the past years' examination true questions are counted first, and the occurrence frequency of the to-be-recommended words in the counted past years' examination true questions is counted after word segmentation processing.

[0082] The word recommendation method provided by the embodiment of the application provides data support for the to-be-recommended words, the occurrence frequency of each word and the memory amount of each word by the user for the subsequent process by constructing the practice question bank through the target question.

[0083] In this embodiment, the prediction model is obtained by the following steps: obtaining a plurality of feedback results and a type of feedback results of a plurality of word practice questions, the type of feedback results including correct use of a word and incorrect use of a word, and the word practice question including a stem and a plurality of options, wherein the stem includes a word or a Chinese interpretation of the word, when the stem is the word, the plurality of options are Chinese interpretations associated with the word, when the stem is the Chinese interpretation of the word, the plurality of options are words associated with the Chinese interpretation, and the association between the plurality of options is determined based on cosine similarity; one-hot encoding the plurality of feedback results to obtain sample data; training the LSTM network with the sample data as the training sample and the type of feedback results as the training label to obtain the prediction model.

[0084] In this embodiment, the word practice question can be a single-choice question or a multiple-choice question.

[0085] In this embodiment, the number of options for each exercise question can be customized according to user needs, for example, the number of options can be 3, 4, 5 or other values.

[0086] In this embodiment, the word exercise question is in the form of an objective question and is presented to the user, and the word proficiency is evaluated by collecting the user's exercise results.

[0087] In this embodiment, the word exercise question is presented in two organizational forms, namely:

[0088] Method 1, the stem of the target question is an English word, and the options are Chinese semantics.

[0089] Word exercise question option organization method: first train a Word2Vec model using a word library, which includes a plurality of English words and a plurality of Chinese characters.

[0090] In this embodiment, the Word2Vec model is used to determine a plurality of Chinese semantic options associated with the English word of the stem, for example, in a multiple-choice question, the stem examines the usage of word A, and the answer options are generally four, one of which is the correct answer, and the other three are the confusion answers, which are obtained by inputting the correct answer into the Word2Vec model for prediction.

[0091] In this embodiment, the relevance of the correct answer and the confusion answer can be evaluated using cosine similarity; for example, the vector representation of word A is [x1, x2,...x n ], the vector representation of word B is [y1, y,...y n ], x1, x2,...x n represent n semantic features of word A, and y1, y,...y n represent n semantic features of word B, then the relevance between word A and word B is determined using the following formula:

[0092]

[0093] In this embodiment, a confusion answer number threshold M is set, and after calculating the M confusion answers with the highest relevance to the correct answer, the M confusion answers and the correct answer are used as candidate answers for the stem to set the target question, and the completion of the user for each target question is counted, that is, the user's error and correct answer.

[0094] When organizing the question, the trained Chinese Word2vec model is used to obtain the Chinese word vector, and the three Chinese words with the closest cosine similarity to the correct option are selected as the options.

[0095] Method 2: The stem is Chinese semantics, and the options are English words.

[0096] In this embodiment, when organizing the target test questions, the word vectors of English words are obtained using the trained English Word2vec model, and the three English words most similar to the correct option are selected as the options by calculating the cosine similarity of the word vectors.

[0097] In this embodiment, the trained LSTM model can be used to predict the user's proficiency in words.

[0098] In this embodiment, the LSTM model is trained by using the historical word practice result data of other users to obtain a prediction model, and then the proficiency of the user in words is predicted; the proficiency of the user in words can be represented by a numerical value, and the value range of the numerical value is [0, 1].

[0099] Figure 2 is the association diagram provided by the application based on the "hidden" state, input vector sequence and output vector sequence of the LSTM model; in Figure 2 The embodiment shown in the figure, the LSTM model maps an input vector sequence x1…x T to an output vector sequence y1…y T This mapping can be determined by calculating a series of "hidden" states h1…h T The "hidden" state can be regarded as a continuous encoding of the relevant information obtained after observing the historical word practice results of the user, and these observation results are also useful for future prediction; wherein, h t (1≤t≤T) is determined as follows:

[0100] h t =tanh(W hx x t +W hh h t-1 +b h );

[0101] W hx is the state-input weight; x t belongs to x1…x T ; W hh is the state-state weight; b h is the bias term of the hidden unit.

[0102] The determination method of the output vector sequence y t is as follows:

[0103] h t =σ(W yx h t +b y );

[0104] wherein, sigma is a sigmoid function; W yx is an information reading weight; b y is a bias term of the information reading unit.

[0105] In this embodiment, the user word exercise feedback result is recorded as 0 (wrong) and 1 (correct), and when training the LSTM model using the historical feedback records (corresponding feedback results) of other user word exercises, first, the historical feedback records of other user word exercises are preprocessed using one-hot encoding, the preprocessed data is used as the input data for training the LSTM model, and the LSTM prediction model is trained and saved.

[0106] In this embodiment, when the system predicts the proficiency of the user's words, first, the trained LSTM model is loaded, and then the historical word exercise feedback records of the user are input into the LSTM model to predict the proficiency of the user for all words.

[0107] The word recommendation method provided in the embodiment of the application uses the multiple feedback results of other user exercises as training samples, trains an LSTM model to obtain a prediction model using the types of feedback results as training labels, realizes the prediction of the proficiency of the user's words, and provides data for subsequent acquisition of the proficiency scores of the user's words.

[0108] In some embodiments, the multiple words to be recommended, the occurrence frequency of each word, and the memory amount of the user for each word are obtained, and the method further includes: determining the memory amount of the user for each word based on the Ebbinghaus forgetting curve.

[0109] In this embodiment, since the memory time interval of the user for the word is positively correlated with the memory amount, the word review plan of the user at different memory time intervals can be determined according to the Ebbinghaus forgetting curve; for example, according to research, the memory amount of the user is 100% when the user just finishes memorizing a word; the memory amount is 58.2% after 20 minutes; the memory amount is 44.2% after 1 hour; the memory amount is 35.8% after 8 to 9 hours; the memory amount is 33.7% after 1 day; the memory amount is 27.8% after 2 days; the memory amount is 25.4% after 6 days; in summary, as the memory time interval of the user for the word increases, the memory amount of the user for the word gradually decreases until completely forgotten.

[0110] In this embodiment, the memory time interval of the user for each word is counted, and the memory amount of the user for each word can be determined according to the Ebbinghaus forgetting curve.

[0111] The word recommendation method provided by the embodiment of the present application can quickly determine the memory amount of each word of the user by setting the word review plan of the user by using the Ebbinghaus forgetting curve, and provide data for subsequent acquisition of the memory amount score of the word of the user.

[0112] In this embodiment, the comprehensive score of each word is determined based on the frequency score of each word, the proficiency score of each word of the user and the memory amount score of each word of the user, comprising: applying the following formula to determine the comprehensive score of each word:

[0113] R u =|I(u)| -0.5 ×∑ v∈I(u) x v ,

[0114] wherein R u is the comprehensive score, I(u) is the implicit feedback of all words of the user u, x v is the implicit feedback feature vector determined based on the frequency score, the proficiency score and the memory amount score, v is a word, and u is a user.

[0115] In this embodiment, after the frequency score P v , the proficiency score D uv and the memory amount score A uv of each word are acquired, the implicit feedback feature vector x v =[P v , D uv , A uv ] is calculated; the implicit feedback I(u) of all words of a certain specific user is set, and the cumulative feature contribution is:

[0116] ∑ v∈I(u) x v ;

[0117] After the above formula is normalized, the comprehensive score is obtained:

[0118] R u =|I(u)| -0.5 ×∑ v∈I(u) x v ;

[0119] According to the comprehensive score, the word with the highest score is selected as the target recommended word and recommended to the user.

[0120] The word recommendation method provided by the embodiment of the present application realizes comprehensive evaluation of the degree of recommendation of a word by using the three dimensions of the occurrence frequency of the word, the proficiency of the user to the word and the memory amount of the user to the word, so as to obtain the amount of words to be recited by the user, and improves the efficiency of the user in reciting words.

[0121] The word recommendation device provided by the present application is described below, and the word recommendation device described below can be referred to in correspondence with the word recommendation method described above.

[0122] Figure 3 The word recommendation device provided by the present application is described below, and the word recommendation device described below can be referred to in correspondence with the word recommendation method described above.

[0123] The word acquisition module 310 is configured to acquire a plurality of words to be recommended, the occurrence frequency of each word and the memory amount of the user to each word.

[0124] The prediction module 320 is configured to determine the proficiency of the user to each word based on a prediction model, wherein the prediction model takes the feedback results of other users using a plurality of word exercises as training samples, and takes the type of the feedback results as training labels.

[0125] The word recommendation module 330 is configured to comprehensively evaluate the plurality of words based on the occurrence frequency, the proficiency and the memory amount of the word, and determine a target recommended word according to the comprehensive evaluation result, wherein the target recommended word is one of the plurality of words.

[0126] The word recommendation device provided by the embodiment of the present application can obtain the amount of words to be recited by the user by predicting the proficiency of the user to each word through the prediction model, comprehensively evaluating each word through the occurrence frequency, the proficiency and the memory amount of the word, and finally determining the target recommended word according to the comprehensive evaluation result, so as to improve the accuracy of word recommendation and further improve the efficiency of the user in reciting words.

[0127] Figure 4 An example of an entity structure schematic diagram of an electronic device is shown as Figure 4As shown, the electronic device can include a processor 410, a communications interface 420, a memory 430, and a communications bus 440, wherein the processor 410, the communications interface 420, and the memory 430 complete mutual communication through the communications bus 440. The processor 410 can invoke a logical instruction in the memory 430 to execute a word recommendation method, which includes: obtaining a plurality of words to be recommended, an occurrence frequency of each word, and a memory amount of each word by a user; determining a proficiency of each word by the user based on a prediction model, the prediction model being obtained with a feedback result of other users using a plurality of word exercises as a training sample and a type of the feedback result as a training label; comprehensively evaluating the plurality of words based on the occurrence frequency, the proficiency, and the word memory amount, and determining a target recommended word according to a comprehensive evaluation result, the target recommended word being one of the plurality of words.

[0128] In addition, the logical instruction in the memory 430 described above can be implemented in the form of a software function unit and sold or used as an independent product, and can be stored in a computer-readable storage medium. Based on such understanding, the technical solutions of the present application essentially or the part that contributes to the prior art or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium, and includes several instructions for making a computer device (which can be a personal computer, a server, or a network device, etc.) execute all or part of the steps of the methods described in the various embodiments of the present application. The aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk or an optical disk, and various media that can store program codes.

[0129] On the other hand, the present application also provides a computer program product, which includes a computer program, the computer program can be stored on a non-transitory computer readable storage medium, and the computer program can be executed by a processor to enable a computer to execute the word recommendation method provided by the above-mentioned methods, which includes: obtaining a plurality of words to be recommended, an occurrence frequency of each word, and a memory amount of each word by a user; determining a proficiency of each word by the user based on a prediction model, the prediction model being obtained with a feedback result of other users using a plurality of word exercises as a training sample and a type of the feedback result as a training label; comprehensively evaluating the plurality of words based on the occurrence frequency, the proficiency, and the word memory amount, and determining a target recommended word according to a comprehensive evaluation result, the target recommended word being one of the plurality of words.

[0130] In yet another aspect, the present application also provides a non-transitory computer readable storage medium having stored thereon a computer program, which, when executed by a processor, implements a word recommendation method provided by each of the above methods, the method comprising: obtaining a plurality of words to be recommended, an occurrence frequency of each word, and a memory amount of each word for a user; determining a proficiency of each word for the user based on a prediction model, the prediction model taking feedback results of other users using a plurality of words for practice as training samples, and taking types of the feedback results as training labels; comprehensively evaluating the plurality of words based on the occurrence frequency, the proficiency, and the memory amount of the word, and determining a target recommended word according to a comprehensive evaluation result, the target recommended word being one of the plurality of words.

[0131] The device embodiments described above are merely illustrative, wherein the units described as separate components can or can not be physically separate, and the components displayed as units can or can not be physical units, i.e., can be located in one place, or can be distributed on multiple network units. Part or all of the modules can be selected to achieve the purpose of the embodiment scheme according to actual needs. Those skilled in the art can understand and implement without creative labor.

[0132] From the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be realized by means of software and necessary general hardware platforms, and of course can also be realized by hardware. Based on such understanding, the above technical solutions can be embodied in the form of software products, which can be stored in a computer readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and include a number of instructions to make a computer device (which can be a personal computer, a server, or a network device, etc.) execute the methods described in each embodiment or some parts of the embodiments.

[0133] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, and not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement for some technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A word recommendation method characterized by, The method comprises the following steps: obtaining a plurality of words to be recommended, the frequency of occurrence of each word, and the memory amount of each word for a user; determining the proficiency of each word for the user based on a prediction model, wherein the prediction model takes feedback results of a plurality of word exercises of other users as training samples, and takes the types of the feedback results as training labels; comprehensively evaluating the plurality of words based on the frequency of occurrence, the proficiency, and the memory amount of each word, and determining a target recommended word according to the comprehensive evaluation result, wherein the target recommended word is one of the plurality of words; the comprehensive evaluation of the plurality of words based on the frequency of occurrence, the proficiency, and the memory amount of each word, and the determination of the target recommended word according to the comprehensive evaluation result, comprises: determining the frequency score of each word based on the mapping relationship between the frequency of occurrence and a preset frequency score weight; determining the proficiency score of each word for the user based on the mapping relationship between the proficiency and a preset proficiency score weight; determining the memory amount of each word for the user based on the mapping relationship between the memory amount and a preset memory amount score weight; determining the comprehensive score of each word based on the frequency score of each word, the proficiency score of each word for the user, and the memory amount score of each word for the user; determining the word with the highest comprehensive score as the target recommended word based on the comprehensive score of each word; the determination of the comprehensive score of each word based on the frequency score of each word, the proficiency score of each word for the user, and the memory amount score of each word for the user, comprises: applying the following formula to determine the comprehensive score of each word: R u = | I(u) | -0.5 x ∑ v∈I(u) x v , wherein R u is the comprehensive score, I(u) is the implicit feedback of all words of user u, I(u) is directly set, x v is an implicit feedback feature vector determined based on the frequency score, the proficiency score and the memory amount score, v is a word, and u is a user; after obtaining the frequency score P v , the proficiency score D uv and the memory amount score A uv corresponding to each word, the implicit feedback feature vector x v =[P v , D uv , A uv ] is calculated.

2. The word recommendation method according to claim 1, characterized in that, the obtaining of the plurality of words to be recommended, the frequency of occurrence of each word, and the memory amount of each word for the user further comprises: obtaining the plurality of words and the frequency of occurrence of each word through word segmentation processing and statistics based on past years' examination true questions.

3. The word recommendation method according to claim 1, characterized in that, the prediction model is obtained through the following steps: obtaining the feedback results of a plurality of word exercises and the types of the plurality of feedback results, wherein the types of the feedback results include correct use of words and incorrect use of words; wherein the word exercise includes a stem and a plurality of options, wherein the stem includes a word or a Chinese interpretation of a word, when the stem is a word, the plurality of options are respectively Chinese interpretations associated with the word, when the stem is a Chinese interpretation of a word, the plurality of options are respectively words associated with the Chinese interpretation, and the association between the plurality of options is determined based on cosine similarity; one-hot encoding the feedback results of the plurality of word exercises to obtain sample data; training an LSTM network with the sample data as training samples and the types of the plurality of feedback results as training labels to obtain the prediction model.

4. The word recommendation method according to claim 1, characterized in that, the obtaining of the plurality of words to be recommended, the frequency of occurrence of each word, and the memory amount of each word for the user further comprises: determining the memory amount of each word for the user based on the Ebbinghaus forgetting curve.

5. A word recommendation device that applies the word recommendation method according to claim 1, characterized by The method comprises the following steps: a word obtaining module is configured to obtain a plurality of words to be recommended, the frequency of occurrence of each word, and the memory amount of each word for a user; The prediction module is configured to determine the proficiency of the user for each word based on a prediction model, wherein the prediction model is trained by using feedback results of other users using a plurality of word exercises as training samples and using types of the feedback results as training labels; The word recommendation module is configured to comprehensively evaluate the plurality of words based on the frequency of occurrence, the proficiency and the memory amount of the words, and determine a target recommended word according to a comprehensive evaluation result, wherein the target recommended word is one of the plurality of words.

6. An electronic device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, The processor implements the word recommendation method according to any one of claims 1 to 4 when executing the program.

7. A non-transitory computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program is configured to implement the word recommendation method according to any one of claims 1 to 4 when executed by the processor.

8. A computer program product comprising a computer program, characterized in that, The computer program is configured to implement the word recommendation method according to any one of claims 1 to 4 when executed by the processor.

Images