A knowledge graph-based intelligent word recommendation method and device for English learning

Abstract

This invention discloses an intelligent recommendation method and apparatus for English learning words based on knowledge graphs, relating to the field of information recommendation. The method includes: accessing a target user's word learning database constructed based on word learning data and a knowledge graph; acquiring multiple words to be learned and multiple learning coefficients from the target user's word learning database, and simultaneously acquiring multiple learning context feature vectors for the multiple words to be learned; calculating the deviation based on multiple standard learning context feature vectors, and correcting based on the multiple learning coefficients to obtain multiple context learning coefficients; performing ranking position analysis and contextual definition recommendation analysis to obtain multiple fused position parameters and multiple contextual definition recommendation coefficient sets; and recommending multiple words to be learned and their corresponding multiple contextual definition sets. This solves the problems of inaccurate word recommendation, poor word recommendation effect, and poor targeting of recommendation results in existing technologies.

Description

Technical Field

[0001] This invention relates to the field of information recommendation, specifically to a method and apparatus for intelligent recommendation of English learning words based on knowledge graphs. Background Technology

[0002] Traditional English vocabulary learning recommendation methods typically only recommend learning volume based on the user's learning progress, lacking the utilization of user vocabulary learning process data. This leads to technical problems such as mismatch between recommendation results and the user's actual situation, inaccurate vocabulary recommendations, poor recommendation effects, and poor targeting of recommendation results. Summary of the Invention

[0003] This invention provides a knowledge graph-based intelligent word recommendation method and device for English learning, which solves the problems of inaccurate word recommendations, poor recommendation effects, and poor targeting of recommendation results in the prior art.

[0004] In view of the above problems, the present invention provides a method and apparatus for intelligent recommendation of English learning words based on knowledge graph.

[0005] In a first aspect, the present invention provides an intelligent recommendation method for English learning words based on knowledge graphs, the method comprising: The target user's word learning database is invoked. The word learning database uses the target user's word learning data from a historical period as the construction data. It is constructed based on a knowledge graph. The word learning data includes words, word accuracy, contextual definitions, and contextual definition accuracy. Obtain multiple words to be learned and multiple learning coefficients from the target user's word learning database, and based on the word learning database, obtain multiple learning context feature vectors for the multiple words to be learned; Obtain multiple standard learning context feature vectors for multiple words to be learned, calculate the deviation between the multiple learning context feature vectors and the multiple standard learning context feature vectors, and correct based on multiple learning coefficients to obtain multiple context learning coefficients; Based on multiple learning coefficients and multiple contextual learning coefficients, ranking position analysis and contextual definition recommendation analysis are performed to obtain multiple fused position parameters and multiple contextual definition recommendation coefficient sets, and recommendations are made for multiple words to be learned and their corresponding multiple contextual definition sets.

[0006] Secondly, the present invention provides an intelligent recommendation device for English learning words based on knowledge graphs, the device comprising: The word learning database construction module is used to call the target user's word learning database. The word learning database uses the target user's word learning data from a historical period as the construction data and is built based on a knowledge graph. The word learning data includes words, word accuracy, contextual definitions, and contextual definition accuracy. The learning context feature vector acquisition module is used to acquire multiple words to be learned and multiple learning coefficients in the target user's word learning database, and based on the word learning database, acquire multiple learning context feature vectors of the multiple words to be learned; The coefficient correction module is used to obtain multiple standard learning context feature vectors for multiple words to be learned, calculate the deviation between the multiple learning context feature vectors and the multiple standard learning context feature vectors, and correct them based on multiple learning coefficients to obtain multiple context learning coefficients. The word recommendation module is used to perform ranking position analysis and contextual definition recommendation analysis based on multiple learning coefficients and multiple contextual learning coefficients, to obtain multiple fused position parameters and multiple contextual definition recommendation coefficient sets, and to recommend multiple words to be learned and their corresponding multiple contextual definition sets.

[0007] One or more technical solutions provided in this invention have at least the following technical effects or advantages: This invention provides a knowledge graph-based intelligent recommendation method and apparatus for English learning words. First, a word learning database for the target user is constructed, and then this database is invoked to provide a multi-dimensional data foundation for subsequent personalized recommendations. Second, multiple words to be learned and their learning coefficients are obtained, and multiple learning context feature vectors are extracted. Third, standard learning context feature vectors are obtained, deviation is calculated, and corrected using the learning coefficients to obtain context learning coefficients, improving the accuracy of word recommendations. Then, based on the learning coefficients and context learning coefficients, ranking position analysis and contextual definition recommendation analysis are performed to obtain a fused position parameter and contextual definition recommendation coefficient set. Finally, words to be learned are ranked and recommended according to the fused position parameters, and multiple contextual definitions corresponding to each word are recommended according to the contextual definition recommendation coefficient set, improving the personalization and learning efficiency of word recommendations. Attached Figure Description

[0008] Figure 1 This is a flowchart illustrating the intelligent recommendation method for English learning words based on knowledge graphs provided by this invention. Figure 2 This is a logical diagram of the knowledge graph-based intelligent word recommendation method for English learning provided by the present invention; Figure 3 This is a schematic diagram of the structure of the knowledge graph-based intelligent word recommendation device for English learning provided by the present invention.

[0009] In the attached diagram, the components represented by each number are as follows: The module includes: 11 Word learning database construction module, 12 Learning context feature vector acquisition module, 13 Coefficient correction module, and 14 Word recommendation module. Detailed Implementation

[0010] This invention provides a knowledge graph-based intelligent word recommendation method and device for English learning, which solves the problems of inaccurate word recommendation, poor word recommendation effect, and poor targeting of recommendation results in the prior art.

[0011] The present invention will now be described in detail with reference to the accompanying drawings.

[0012] Example 1, as Figure 1 , Figure 2 As shown, this invention provides an intelligent recommendation method for English learning words based on knowledge graphs, the method comprising: S10: Call the target user's word learning database, wherein the word learning database uses the target user's word learning data over a historical period as the construction data, is constructed based on a knowledge graph, and the word learning data includes words, word accuracy, contextual definitions, and contextual definition accuracy. In this embodiment of the invention, contextual interpretation refers to the different meanings of a word in different contexts; word accuracy is the ratio of the number of correct word learning attempts to the total number of learning tests; contextual interpretation accuracy is the ratio of the number of correct word learning attempts for different contextual interpretations to the total number of learning tests.

[0013] Specifically, the target user's historical word learning data is used as the construction data to build a word learning database based on a knowledge graph. Then, the word learning database is called to obtain the words to be learned and their learning coefficients, providing basic data for subsequent word recommendations.

[0014] In the method of this embodiment of the invention, the step of constructing the word learning database includes: Acquire word learning data for multiple words learned by the target user within a historical period. Each word learning data includes the word, word accuracy rate, and contextual interpretation accuracy rate for multiple contextual interpretations. Using words as entities and word accuracy and contextual interpretation as connecting entities, a single-layer word learning database is constructed based on knowledge graphs, employing multiple word learning data. Using the contextual definition accuracy of each word as a two-layer connection entity, multiple contextual definitions are connected to construct a complete word learning database, which is then updated.

[0015] In this embodiment of the invention, firstly, word learning data of multiple words learned by the target user over a historical period of, for example, the past one or two academic years is obtained. Each word learning data includes the word, the word accuracy rate, and the contextual interpretation accuracy rate of multiple contextual interpretations.

[0016] Secondly, words are treated as entities, and word accuracy and contextual meaning are treated as a layer of connecting entities. Each word corresponds to two connecting entities: word accuracy and contextual meaning. These two connecting entities are on the same layer. Then, multiple word learning data are used to connect all words according to their connection methods to build a layer of word learning database.

[0017] Finally, using the contextual definition accuracy rate of each word as a two-layer connection entity, multiple contextual definitions are connected. Each contextual definition corresponds to multiple contextual definition accuracy rates, resulting in the correspondence between words and entities to the two-layer connection entities: User → Word → Word Accuracy Rate, Word → Contextual Definition → Contextual Definition Accuracy Rate. All words are connected according to their connection methods, and a complete word learning database is constructed by combining all words. This database can be updated based on the user's new learning progress. Whenever a user completes a new word exercise, the overall accuracy rate and the accuracy rate of each contextual definition for that word are calculated, and the corresponding numerical entities are updated in the knowledge graph.

[0018] In this embodiment of the invention, word learning data of multiple words learned by a target user within a historical period are obtained, and a first-layer word learning database and a second-layer word learning database are constructed to obtain a complete word learning database, providing a sufficient data foundation for subsequent word learning recommendations.

[0019] S20: Obtain multiple words to be learned and multiple learning coefficients from the target user's word learning database, and based on the word learning database, obtain multiple learning context feature vectors for the multiple words to be learned; In this embodiment of the invention, multiple words to be learned are obtained from the target user's word learning database, and multiple learning coefficients of the word learning database are calculated. The learning coefficients are used to measure the degree to which the target user needs to learn. Then, based on the word learning database, a learning context feature vector is constructed to facilitate subsequent mathematical operations.

[0020] S20 in the method of this embodiment includes: Words with a correct accuracy rate lower than the correct accuracy rate threshold in the word learning database are selected as multiple words to be learned, and multiple negatively correlated learning coefficients are calculated based on the correct accuracy rates of the corresponding multiple words. Based on the word learning database, multiple learning context feature vectors of multiple words to be learned are obtained.

[0021] In this embodiment of the invention, firstly, words with accuracy rates below a threshold in the word learning database are selected as multiple words to be learned. The word accuracy threshold measures the number of words the target user needs to learn and can be dynamically adjusted based on the target user's expected mastery of the words, their historical accuracy rates, and recommendation load requirements. The value range is 0-1, for example, 0.8. This threshold ensures the rationality and stability of the word selection process. Simultaneously, based on the multiple word accuracy rates, several negatively correlated learning coefficients are calculated. The learning coefficient is the difference between 1 and the word accuracy rate, i.e., learning coefficient = 1 - word accuracy rate. The word accuracy rate and learning coefficient are negatively correlated; the lower the word accuracy rate, the greater the learning coefficient the target user needs to learn, and the higher the priority for subsequent recommendations.

[0022] Secondly, based on the word learning database, a learning context feature vector is constructed using the accuracy rates of multiple contextual definitions for each word as vector elements. The number of vector elements is determined by the category of the contextual definition, with one vector element corresponding to each contextual definition category. This ensures that the learning context feature vector is calculated specifically for the words to be learned.

[0023] In the method of this embodiment of the invention, based on the word learning database, multiple learning context feature vectors of multiple words to be learned are obtained, including: Extract the accuracy sets of multiple contextual definitions for multiple words to be learned; Based on multiple contextual interpretation accuracy sets, multiple learning context feature vectors are constructed, wherein the vector dimension of the learning context feature vector is the same as the number of contextual interpretation categories.

[0024] In this embodiment of the invention, firstly, multiple contextual definition accuracy sets are extracted for multiple words to be learned. The contextual definition accuracy set is the collection of all contextual definition accuracy rates for a word, encompassing all contextual categories of the word.

[0025] Secondly, the accuracy rate of each scenario interpretation in the multiple scenario interpretation accuracy sets is used as the corresponding vector element of the scenario feature vector to construct multiple learning scenario feature vectors. The vector dimension of the learning scenario feature vector is the same as the number of scenarios interpretation categories.

[0026] In this embodiment of the invention, multiple words to be learned are obtained by filtering through word accuracy thresholds. Based on the accuracy of multiple words, multiple learning coefficients are calculated to accurately obtain the basic data for subsequent calculations. The contextual interpretation accuracy set is transformed into a learning context feature vector with a fixed order, which facilitates subsequent numerical calculations and quantitative analysis.

[0027] S30: Obtain multiple standard learning context feature vectors for multiple words to be learned, calculate the deviation between the multiple learning context feature vectors and the multiple standard learning context feature vectors respectively, and correct them based on multiple learning coefficients to obtain multiple context learning coefficients; In this embodiment of the invention, the standard learning context feature vector is a feature vector that serves as a benchmark; the deviation is a numerical value that measures the degree of difference between the actual learning context feature vector and the standard vector; and the correction is a process of adjusting the deviation using the learning coefficient to eliminate system bias.

[0028] Specifically, multiple standard learning context feature vectors for multiple words to be learned are obtained, and the deviation between the multiple learning context feature vectors and the multiple standard learning context feature vectors is calculated. Then, based on multiple learning coefficients, the basic context learning coefficients are corrected to obtain multiple context learning coefficients. The context learning coefficients comprehensively reflect the overall mastery of words and the degree of difference within the context.

[0029] S30 in the method of this embodiment includes: Obtain the standard accuracy rate of word learning, and construct multiple standard learning context feature vectors by combining the number of contextual definition categories of multiple words to be learned; Based on Euclidean distance, the deviation between multiple learning context feature vectors and multiple standard learning context feature vectors is calculated to obtain multiple basic context learning coefficients. Based on multiple learning coefficients, the learning coefficients of multiple basic scenarios are corrected and calculated to obtain multiple scenario learning coefficients.

[0030] In this embodiment of the invention, firstly, the standard accuracy rate of word learning is obtained. Then, combined with the number of contextual definition categories for multiple words to be learned, multiple standard learning contextual feature vectors are constructed. The standard accuracy rate is a preset target accuracy value, representing the ideal level of mastery the system expects the target user to achieve at this learning stage. It typically ranges from 0.8 to 1.0, for example, 0.8. It can be set according to the user's learning stage, word difficulty, or system strategy.

[0031] Specifically, using the standard accuracy threshold of words as the element value of each dimension, a vector with the same dimensions as the feature vector of the learning context is constructed to achieve adaptive generation of the standard vector under heterogeneous dimensions. Then, the deviation between the actual learning context feature vector and the standard vector is calculated and normalized to obtain a dimensionless and standardized deviation, eliminating the influence of dimensional differences on the measurement scale and achieving a unified deviation measurement scale.

[0032] The standard learning context feature vector is a feature vector with the same dimension as the learning context feature vector, describing the standard accuracy. The standard accuracy of words corresponding to the feature vectors of successfully learned learning contexts is labeled, and the learning context feature vectors with the labeled standard accuracy of words are used as the standard learning context feature vector. For example, assuming a word has three contextual definitions, the learning context feature vector labeled with a standard accuracy of 80% is used as the standard learning context feature vector, then the standard learning context feature vector is (0.8, 0.8, 0.8). Each word learning session requires understanding multiple contextual definitions; the user must complete all of them to finish the word learning session, thus obtaining the standard accuracy of the word for each contextual definition after that session.

[0033] Secondly, based on Euclidean distance, the deviation between multiple learning context feature vectors and multiple standard learning context feature vectors is calculated. Euclidean distance measures the linear distance between two vectors; the larger the distance, the less similar the two vectors are. Then, each calculated Euclidean distance is normalized, and the normalized Euclidean distance is used as the base context learning coefficient. Normalized Euclidean distance = Euclidean distance / square root of the vector dimension of the learning context feature vector or the standard learning context feature vector. , where d represents the Euclidean distance and n represents the vector dimension of the learning context feature vector or the standard learning context feature vector.

[0034] Specifically, the Euclidean distance formula is used to calculate the Euclidean distance between the learning context feature vectors and multiple standard learning context feature vectors. The Euclidean distance is the difference between corresponding elements of the two vectors. The square root of the sum of the squares of these differences is then taken. The result of the Euclidean distance calculation is used as the deviation between the multiple learning context feature vectors and the multiple standard learning context feature vectors.

[0035] Finally, based on multiple learning coefficients, the learning coefficients for multiple basic contexts are corrected to eliminate the impact of varying word accuracy on context deviation. After correction, multiple context learning coefficients are obtained. These context learning coefficients comprehensively reflect the overall weakness of words and the degree of difference within contexts, and are used for subsequent ranking and recommendation.

[0036] In the method of this embodiment of the invention, multiple basic context learning coefficients are corrected and calculated based on multiple learning coefficients to obtain multiple context learning coefficients, including: Calculate the ratio of each learning coefficient to the mean of multiple learning coefficients to obtain multiple context correction coefficients; Multiple contextual correction coefficients are used to correct and calculate the learning coefficients of multiple basic contexts, resulting in multiple contextual learning coefficients.

[0037] In this embodiment of the invention, firstly, the ratio of each learning coefficient to the mean of multiple learning coefficients is calculated to obtain multiple context correction coefficients. These context correction coefficients are used to adjust the magnitude of the base context learning coefficients. Specifically, the context correction coefficient is calculated as: learning coefficient / mean learning coefficient.

[0038] Secondly, multiple contextual correction coefficients are used to correct and calculate the learning coefficients of multiple basic contexts, resulting in multiple contextual learning coefficients. The contextual learning coefficient is calculated as: Contextual Correction Coefficient × Basic Contextual Learning Coefficient.

[0039] In this embodiment of the invention, a standard accuracy rate for word learning is obtained, and multiple standard learning context feature vectors are constructed by combining the number of contextual definition categories of multiple words to be learned. Then, the deviation between the two vectors is calculated based on Euclidean distance to obtain multiple basic context learning coefficients. Finally, the multiple basic context learning coefficients are corrected based on these multiple learning coefficients to obtain multiple context learning coefficients. Through adaptive correction, the current contextual learning quality of the target user is evaluated to ensure the accuracy and reliability of subsequent recommendations.

[0040] S40: Based on multiple learning coefficients and multiple contextual learning coefficients, perform ranking position analysis and contextual interpretation recommendation analysis to obtain multiple fusion position parameters and multiple contextual interpretation recommendation coefficient sets, and recommend multiple words to be learned and their corresponding multiple contextual interpretation sets.

[0041] In this embodiment of the invention, the ranking position analysis is to rank words by combining indicators such as the learning coefficient and the contextual learning coefficient, and determine the position of each word in the recommendation list; the fusion position parameter is a comprehensive ranking used to finally determine the recommendation order of words; the contextual paraphrase recommendation analysis is to calculate the probability of being recommended and learned for different contextual paraphrases under the same word based on their accuracy or other features.

[0042] Specifically, based on the learning coefficient and multiple contextual learning coefficients, each word is ranked according to the learning consistency parameter of different contextual interpretations. Then, contextual interpretation recommendation analysis is performed to obtain contextual interpretation recommendation coefficients. Word recommendations are made through two dimensions: ranking position analysis and contextual interpretation recommendation analysis, thereby improving the recommendation quality.

[0043] S40 in the method of this embodiment includes: Based on the elements within the feature vector of each learning context, multiple context learning consistency parameters are calculated, and the multiple words to be learned are arranged in ascending order to obtain the context word sequence. Based on multiple learning coefficients and multiple contextual learning coefficients, multiple comprehensive learning coefficients are calculated. Multiple words to be learned are arranged in descending order to obtain a sequence of learned words. Combined with the contextual word sequence, multiple fusion position parameters of multiple words to be learned are calculated. Based on the accuracy of multiple contextual definitions for each word to be learned, a set of recommendation coefficients for multiple contextual definitions is calculated. Based on multiple fusion position parameters, multiple words to be learned are recommended, and based on multiple contextual definition recommendation coefficient sets, multiple contextual definitions for each word to be learned are recommended.

[0044] In this embodiment of the invention, firstly, multiple context learning consistency parameters are calculated based on the elements in the feature vector of each learning context. Then, the multiple words to be learned are arranged in ascending order of the context learning consistency parameters to obtain a context word sequence.

[0045] Secondly, based on multiple learning coefficients and multiple contextual learning coefficients, the comprehensive learning coefficient of each word to be learned is calculated. Then, the multiple words to be learned are arranged in descending order of comprehensive learning coefficients to obtain a sequence of learned words. The sequence of learned words is combined with the sequence of contextual words to calculate multiple fusion position parameters of the multiple words to be learned.

[0046] Next, based on the accuracy rates of multiple contextual definitions for each word to be learned, multiple contextual definition recommendation coefficient sets are calculated. Wherein, the contextual definition recommendation coefficient = 1 - contextual definition accuracy rate.

[0047] Finally, multiple words to be learned are recommended in ascending order of their fusion position parameters; the earlier a word appears in the fusion position parameter list, the higher its priority. Simultaneously, contextual definitions within the contextual definition recommendation coefficient set are sorted in descending order of their coefficients, forming the contextual definition recommendation order for each word. A higher contextual definition recommendation coefficient results in higher priority display; a lower coefficient reduces the likelihood of that definition being displayed first. When a target user learns the word, the word and its contextual definitions are recommended sequentially based on the fusion position parameter and the contextual definition recommendation coefficient.

[0048] In the method of this embodiment of the invention, multiple context learning consistency parameters are calculated based on the elements in the feature vector of each learning context, and multiple words to be learned are arranged in ascending order to obtain a context word sequence, including: Based on multiple elements within the feature vector of each learning scenario, element consistency calculation is performed to obtain multiple scenario learning consistency parameters; By arranging multiple words to be learned in ascending order based on multiple contextual learning consistency parameters, a contextual word sequence is obtained.

[0049] In this embodiment of the invention, firstly, element consistency calculation is performed based on multiple elements within the feature vector of each learning context.

[0050] Specifically, the variance of the learning context feature vector is obtained through statistical analysis. Where n is the number of elements in the learning context feature vector, x is an element of the learning context feature vector, and xˈ is the mean of the elements in the learning context feature vector. Then, the reciprocal of the variance is... This serves as a contextual learning consistency parameter. The value range of this parameter is, for example, 0-2. A larger value indicates that the accuracy rates of interpretations across different contexts are closer, and the target user's understanding of the word is more consistent across contexts. A smaller value indicates more significant differences in the accuracy rates of interpretations across different contexts, suggesting obvious weaknesses in the context. To avoid the contextual learning consistency parameter becoming infinitely large, when the variance is 0, the parameter is set to 2. This effectively constrains the value range of the contextual learning consistency parameter, ensuring the stability of subsequent numerical calculations.

[0051] Secondly, multiple contextual learning consistency parameters are arranged in ascending order to obtain a contextual word sequence. Words with poorer contextual semantic learning consistency are ranked higher and given priority for learning.

[0052] In the method of this embodiment of the invention, multiple words to be learned are arranged in descending order based on multiple learning coefficients and multiple contextual learning coefficients to obtain a sequence of learned words. Combined with the contextual word sequence, multiple fusion position parameters of the multiple words to be learned are calculated, including: Based on multiple learning coefficients and multiple contextual learning coefficients, multiple comprehensive learning coefficients are calculated, and multiple words to be learned are arranged in descending order to obtain a sequence of words to be learned. Obtain the position of each word to be learned in the context word sequence and the learning word sequence, and calculate multiple fusion position parameters.

[0053] In this embodiment of the invention, firstly, multiple comprehensive learning coefficients are calculated based on multiple learning coefficients and multiple contextual learning coefficients. The comprehensive learning coefficient = (learning coefficient + contextual learning coefficient) / 2, which comprehensively reflects the urgency of word learning. The words to be learned are then arranged in descending order of the comprehensive learning coefficients to obtain a learning word sequence. Words appearing earlier in the learning word sequence have lower accuracy rates, higher comprehensive learning coefficients, higher urgency of learning, and a greater likelihood of being recommended for further learning.

[0054] Secondly, the position of each word to be learned within the context word sequence and the learning word sequence is obtained. The average of these positions is then calculated to obtain multiple fusion position parameters. Words are recommended according to these fusion position parameters in ascending order; the smaller the value, the higher the priority for recommendation. For example, if a word to be learned is positioned 1 in the context word sequence and 2 in the learning word sequence, the average value would be 1.5.

[0055] In this embodiment of the invention, the contextual interpretation recommendation coefficient is calculated based on the contextual interpretation accuracy rate; the fusion position parameter is calculated based on the learning coefficient, the contextual learning coefficient, and the contextual internal consistency parameter to make the recommendation results more scientific. Then, recommendations are made sequentially according to the magnitude of the fusion position parameter and the contextual interpretation recommendation coefficient to obtain the final recommendation result, thereby realizing personalized recommendations, improving the recommendation effect, and ensuring the learning effect of the target users.

[0056] Through the above specific implementation methods, the embodiments of the present invention achieve the following technical effects: In this embodiment of the invention, firstly, word learning data of multiple words learned by the target user within a historical period is obtained, and a first-layer word learning database and a second-layer word learning database are constructed to obtain a complete word learning database, which provides a sufficient data foundation for subsequent word learning recommendations.

[0057] Secondly, multiple words to be learned are obtained by filtering through word accuracy thresholds. Based on the accuracy of multiple words, multiple learning coefficients are calculated to accurately obtain the basic data for subsequent calculations. The set of contextual interpretation accuracy is transformed into a learning context feature vector with a fixed order, which facilitates subsequent numerical calculations and quantitative analysis.

[0058] Next, the standard accuracy rate of word learning is obtained, and multiple standard learning context feature vectors are constructed by combining the number of contextual definition categories of multiple words to be learned. Then, the deviation between the two vectors is calculated based on the Euclidean distance to obtain multiple basic context learning coefficients. Finally, the multiple basic context learning coefficients are corrected based on the multiple learning coefficients to obtain multiple context learning coefficients. Through adaptive correction, the current context learning quality of the target user is evaluated to ensure the accuracy and reliability of subsequent recommendations.

[0059] Finally, the contextual interpretation recommendation coefficient is calculated based on the contextual interpretation accuracy rate; the fusion position parameter is calculated based on the learning coefficient, contextual learning coefficient, and contextual internal consistency parameter to make the recommendation results more scientific. Then, recommendations are made sequentially according to the magnitude of the fusion position parameter and the contextual interpretation recommendation coefficient to obtain the final recommendation result, realize the personalization of recommendations, improve the recommendation effect, and ensure the learning effect of the target users.

[0060] Example 2, as Figure 3 As shown, based on the same inventive concept as the knowledge graph-based intelligent recommendation method for English learning words provided in Embodiment 1, this embodiment of the invention also provides a knowledge graph-based intelligent recommendation device for English learning words, the device comprising: The word learning database construction module 11 is used to call the target user's word learning database. The word learning database uses the target user's word learning data from a historical period as the construction data and is constructed based on a knowledge graph. The word learning data includes words, word accuracy, contextual definitions, and contextual definition accuracy. The learning context feature vector acquisition module 12 is used to acquire multiple words to be learned and multiple learning coefficients in the target user's word learning database, and based on the word learning database, acquire multiple learning context feature vectors of the multiple words to be learned. The coefficient correction module 13 is used to obtain multiple standard learning context feature vectors of multiple words to be learned, calculate the deviation between the multiple learning context feature vectors and the multiple standard learning context feature vectors respectively, and correct based on multiple learning coefficients to obtain multiple context learning coefficients. The word recommendation module 14 is used to perform ranking position analysis and contextual definition recommendation analysis based on multiple learning coefficients and multiple contextual learning coefficients, to obtain multiple fused position parameters and multiple contextual definition recommendation coefficient sets, and to recommend multiple words to be learned and their corresponding multiple contextual definition sets.

[0061] In one embodiment, the word learning database construction module 11 is used for: Acquire word learning data for multiple words learned by the target user within a historical period. Each word learning data includes the word, word accuracy, and contextual interpretation accuracy for multiple contextual interpretations. Using words as entities and word accuracy and contextual interpretation as connecting entities, a single-layer word learning database is constructed based on knowledge graphs, employing multiple word learning data. Using the contextual definition accuracy of each word as a two-layer connection entity, multiple contextual definitions are connected to construct a complete word learning database, which is then updated.

[0062] In one embodiment, the learning context feature vector acquisition module 12 is used for: Words with a correct accuracy rate lower than the correct accuracy rate threshold in the word learning database are selected as multiple words to be learned, and multiple negatively correlated learning coefficients are calculated based on the correct accuracy rates of the corresponding multiple words. Based on the word learning database, multiple learning context feature vectors of multiple words to be learned are obtained.

[0063] Specifically, based on the word learning database, multiple learning context feature vectors for multiple words to be learned are obtained, including: Extract the accuracy sets of multiple contextual definitions for multiple words to be learned; Based on multiple contextual interpretation accuracy sets, multiple learning context feature vectors are constructed, wherein the vector dimension of the learning context feature vector is the same as the number of contextual interpretation categories.

[0064] In one embodiment, the coefficient correction module 13 is used for: Obtain the standard accuracy rate of word learning, and construct multiple standard learning context feature vectors by combining the number of contextual definition categories of multiple words to be learned; Based on Euclidean distance, the deviation between multiple learning context feature vectors and multiple standard learning context feature vectors is calculated to obtain multiple basic context learning coefficients. Based on multiple learning coefficients, the learning coefficients of multiple basic scenarios are corrected and calculated to obtain multiple scenario learning coefficients.

[0065] Among them, based on multiple learning coefficients, multiple basic context learning coefficients are corrected and calculated to obtain multiple context learning coefficients, including: Calculate the ratio of each learning coefficient to the mean of multiple learning coefficients to obtain multiple context correction coefficients; Multiple contextual correction coefficients are used to correct and calculate multiple contextual learning coefficients, resulting in multiple contextual learning coefficients.

[0066] In one embodiment, the word recommendation module 14 is used for: Based on the elements within the feature vector of each learning context, multiple context learning consistency parameters are calculated, and the multiple words to be learned are arranged in ascending order to obtain the context word sequence. Based on multiple learning coefficients and multiple contextual learning coefficients, multiple comprehensive learning coefficients are calculated. Multiple words to be learned are arranged in descending order to obtain a sequence of learned words. Combined with the contextual word sequence, multiple fusion position parameters of multiple words to be learned are calculated. Based on the accuracy of multiple contextual definitions for each word to be learned, a set of recommendation coefficients for multiple contextual definitions is calculated. Based on multiple fusion position parameters, multiple words to be learned are recommended, and based on multiple contextual definition recommendation coefficient sets, multiple contextual definitions for each word to be learned are recommended.

[0067] Specifically, based on the elements within the feature vector of each learning context, multiple context learning consistency parameters are calculated. These parameters are then arranged in ascending order of the words to be learned, resulting in a context word sequence, including: Based on multiple elements within the feature vector of each learning scenario, element consistency calculation is performed to obtain multiple scenario learning consistency parameters; By arranging multiple words to be learned in ascending order based on multiple contextual learning consistency parameters, a contextual word sequence is obtained.

[0068] Specifically, based on multiple learning coefficients and multiple contextual learning coefficients, multiple words to be learned are arranged in descending order to obtain a sequence of learned words. Combined with the contextual word sequence, multiple fusion position parameters of the multiple words to be learned are calculated, including: Based on multiple learning coefficients and multiple contextual learning coefficients, multiple comprehensive learning coefficients are calculated, and multiple words to be learned are arranged in descending order to obtain a sequence of words to be learned. Obtain the position of each word to be learned in the context word sequence and the learning word sequence, and calculate multiple fusion position parameters.

[0069] Compared to existing technologies, this invention provides a knowledge graph-based intelligent recommendation method and device for English learning words. It improves the construction of a word learning database, the acquisition of learning context feature vectors, coefficient correction, and word recommendation, ultimately yielding personalized recommendation results for target users. First, it obtains word learning data for multiple words, forming a word learning database. Second, it acquires multiple words to be learned, calculates multiple learning coefficients, and obtains learning context feature vectors, providing foundational data for subsequent numerical calculations. Third, it constructs standard learning context feature vectors, calculates basic context learning coefficients, and corrects these coefficients to obtain context learning coefficients. Adaptive correction ensures the accuracy and reliability of subsequent recommendations. Finally, it calculates contextual definition recommendation coefficients and fusion position parameters, and recommends words sequentially according to the magnitude of these coefficients, obtaining the final recommendation result. This achieves personalized recommendations, improves recommendation effectiveness, and ensures the learning outcomes for target users.

[0070] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A knowledge graph-based intelligent word recommendation method for English learning, characterized in that, The method includes: The target user's word learning database is invoked. The word learning database uses the target user's word learning data from a historical period as the construction data. It is constructed based on a knowledge graph. The word learning data includes words, word accuracy, contextual definitions, and contextual definition accuracy. Obtain multiple words to be learned and multiple learning coefficients from the target user's word learning database, and based on the word learning database, obtain multiple learning context feature vectors for the multiple words to be learned; Obtain multiple standard learning context feature vectors for multiple words to be learned, calculate the deviation between the multiple learning context feature vectors and the multiple standard learning context feature vectors, and correct based on multiple learning coefficients to obtain multiple context learning coefficients; Based on multiple learning coefficients and multiple contextual learning coefficients, ranking position analysis and contextual definition recommendation analysis are performed to obtain multiple fused position parameters and multiple contextual definition recommendation coefficient sets, and recommendations are made for multiple words to be learned and their corresponding multiple contextual definition sets.

2. The intelligent recommendation method for English learning words based on knowledge graphs according to claim 1, characterized in that, The steps for constructing the word learning database include: Acquire word learning data for multiple words learned by the target user within a historical period. Each word learning data includes the word, word accuracy, and contextual interpretation accuracy for multiple contextual interpretations. Using words as entities and word accuracy and contextual interpretation as connecting entities, a single-layer word learning database is constructed based on knowledge graphs, employing multiple word learning data. Using the contextual definition accuracy of each word as a two-layer connection entity, multiple contextual definitions are connected to construct a complete word learning database, which is then updated.

3. The intelligent recommendation method for English learning words based on knowledge graphs according to claim 1, characterized in that, Obtain multiple words to be learned and multiple learning coefficients from the target user's word learning database, and based on the word learning database, obtain multiple learning context feature vectors for the multiple words to be learned, including: Words with a correct accuracy rate lower than the correct accuracy rate threshold in the word learning database are selected as multiple words to be learned, and multiple negatively correlated learning coefficients are calculated based on the correct accuracy rates of the corresponding multiple words. Based on the word learning database, multiple learning context feature vectors of multiple words to be learned are obtained.

4. The intelligent recommendation method for English learning words based on knowledge graphs according to claim 3, characterized in that, Based on the aforementioned word learning database, multiple learning context feature vectors for multiple words to be learned are obtained, including: Extract the accuracy sets of multiple contextual definitions for multiple words to be learned; Based on multiple contextual interpretation accuracy sets, multiple learning context feature vectors are constructed, wherein the vector dimension of the learning context feature vector is the same as the number of contextual interpretation categories.

5. The intelligent recommendation method for English learning words based on knowledge graphs according to claim 1, characterized in that, Obtain multiple standard learning context feature vectors for multiple words to be learned, calculate the deviation between each learning context feature vector and the multiple standard learning context feature vectors, and correct based on multiple learning coefficients to obtain multiple context learning coefficients, including: Obtain the standard accuracy rate of word learning, and construct multiple standard learning context feature vectors by combining the number of contextual definition categories of multiple words to be learned; Based on Euclidean distance, the deviation between multiple learning context feature vectors and multiple standard learning context feature vectors is calculated to obtain multiple basic context learning coefficients. Based on multiple learning coefficients, the learning coefficients of multiple basic scenarios are corrected and calculated to obtain multiple scenario learning coefficients.

6. The intelligent recommendation method for English learning words based on knowledge graphs according to claim 5, characterized in that, Based on multiple learning coefficients, correction calculations are performed on multiple basic context learning coefficients to obtain multiple context learning coefficients, including: Calculate the ratio of each learning coefficient to the mean of multiple learning coefficients to obtain multiple context correction coefficients; Multiple contextual correction coefficients are used to correct and calculate the learning coefficients of multiple basic contexts, resulting in multiple contextual learning coefficients.

7. The intelligent recommendation method for English learning words based on knowledge graphs according to claim 1, characterized in that, Based on multiple learning coefficients and multiple contextual learning coefficients, ranking position analysis and contextual definition recommendation analysis are performed to obtain multiple fused position parameters and multiple contextual definition recommendation coefficient sets. Recommendations are then made for multiple words to be learned and their corresponding multiple contextual definition sets, including: Based on the elements within the feature vector of each learning context, multiple context learning consistency parameters are calculated, and the multiple words to be learned are arranged in ascending order to obtain the context word sequence. Based on multiple learning coefficients and multiple contextual learning coefficients, multiple comprehensive learning coefficients are calculated. Multiple words to be learned are arranged in descending order to obtain a sequence of learned words. Combined with the contextual word sequence, multiple fusion position parameters of multiple words to be learned are calculated. Based on the accuracy of multiple contextual definitions for each word to be learned, a set of recommendation coefficients for multiple contextual definitions is calculated. Based on multiple fusion position parameters, multiple words to be learned are recommended, and based on multiple contextual definition recommendation coefficient sets, multiple contextual definitions for each word to be learned are recommended.

8. The intelligent recommendation method for English learning words based on knowledge graphs according to claim 7, characterized in that, Based on the elements within the feature vector of each learning context, multiple context learning consistency parameters are calculated. These parameters are then arranged in ascending order of the words to be learned, resulting in a context word sequence, including: Based on multiple elements within the feature vector of each learning scenario, element consistency calculation is performed to obtain multiple scenario learning consistency parameters; By arranging multiple words to be learned in ascending order based on multiple contextual learning consistency parameters, a contextual word sequence is obtained.

9. The intelligent recommendation method for English learning words based on knowledge graphs according to claim 7, characterized in that, Based on multiple learning coefficients and multiple contextual learning coefficients, multiple comprehensive learning coefficients are calculated. The multiple words to be learned are then arranged in descending order to obtain a sequence of learned words. Combined with the contextual word sequence, multiple fusion position parameters of the multiple words to be learned are calculated, including: Based on multiple learning coefficients and multiple contextual learning coefficients, multiple comprehensive learning coefficients are calculated, and multiple words to be learned are arranged in descending order to obtain a sequence of words to be learned. Obtain the position of each word to be learned in the context word sequence and the learning word sequence, and calculate multiple fusion position parameters.

10. A knowledge graph-based intelligent word recommendation device for English learning, characterized in that, The apparatus for implementing the knowledge graph-based intelligent word recommendation method for English learning according to any one of claims 1-9, the apparatus comprising: The word learning database construction module is used to call the target user's word learning database. The word learning database uses the target user's word learning data from a historical period as the construction data and is built based on a knowledge graph. The word learning data includes words, word accuracy, contextual definitions, and contextual definition accuracy. The learning context feature vector acquisition module is used to acquire multiple words to be learned and multiple learning coefficients in the target user's word learning database, and based on the word learning database, acquire multiple learning context feature vectors of the multiple words to be learned; The coefficient correction module is used to obtain multiple standard learning context feature vectors for multiple words to be learned, calculate the deviation between the multiple learning context feature vectors and the multiple standard learning context feature vectors, and correct them based on multiple learning coefficients to obtain multiple context learning coefficients. The word recommendation module is used to perform ranking position analysis and contextual definition recommendation analysis based on multiple learning coefficients and multiple contextual learning coefficients, to obtain multiple fused position parameters and multiple contextual definition recommendation coefficient sets, and to recommend multiple words to be learned and their corresponding multiple contextual definition sets.

Images