Assistant robot processing system based on natural language processing

Abstract

The invention discloses an assistant robot processing system based on natural language processing, which belongs to the technical field of robot processing and comprises a knowledge point text acquisition module, a knowledge point text preprocessing module, a knowledge point text vectorization processing module, a knowledge point text clustering module and an assistant robot processing module. The method specifically comprises the following steps: introducing a chaos adaptive walk strategy to improve a dolphin group algorithm to obtain part-of-speech influence, obtaining a distribution concentration degree based on distribution entropy, generating a word importance feature vector, and splicing the word importance feature vector with a semantic feature vector to obtain a knowledge point text vector; initializing a clustering center and a dimension weight, constructing a membership matrix according to a double-feature weighted distance, updating the clustering center according to a credibility index, combining a clustering contribution degree, a part-of-speech influence, a distribution concentration degree and a variance to update the dimension weight, judging convergence and obtaining a clustering result, so that a recommendation result of the teaching assistant robot stably focuses on a core knowledge point; and the recommendation accuracy of the teaching assistant robot is improved.

Description

Technical Field

[0001] The present invention belongs to the field of robot processing technology, and specifically refers to a teaching assistant robot processing system based on natural language processing. Background Art

[0002] A teaching assistant robot processing system utilizes natural language processing and machine learning technologies to intelligently respond to user learning questions, quickly match the most relevant knowledge points, and generate personalized recommendation lists, thereby providing users with accurate and efficient learning assistance. However, existing teaching assistant robot processing systems rely solely on word frequency-inverse document frequency or semantic vectors, failing to accurately reflect the semantics and core information of knowledge points. This makes it difficult for users to find relevant knowledge points when performing query matching, affecting the accuracy of the teaching assistant robot's recommendations. Existing teaching assistant robot processing systems also fail to automatically identify and highlight truly core knowledge points, are easily distracted by noisy text, and lack dynamic adaptation to the importance of information across different dimensions. As a result, the teaching assistant robot's recommendations often contain marginal or irrelevant content, making it difficult for users to quickly identify the desired knowledge points. Summary of the Invention

[0003] In view of the above situation, in order to overcome the defects of the prior art, the present invention provides a teaching assistant robot processing system based on natural language processing. In view of the problem that the existing teaching assistant robot processing system only relies on word frequency-inverse document frequency or only relies on semantic vectors, which cannot accurately reflect the semantics and core information of knowledge points, resulting in difficulty in finding relevant knowledge points when users query and match, affecting the recommendation accuracy of the teaching assistant robot, this solution calculates word frequency and inverse document frequency, introduces a chaotic adaptive wandering strategy to improve the dolphin group algorithm, obtains the part-of-speech influence, obtains the distribution concentration based on the distribution entropy, generates a word importance feature vector, and splices it with the semantic feature vector to obtain a knowledge point text vector. The generated vector takes into account both vocabulary importance and deep semantics, accurately depicts the core features of the knowledge point, provides a high-quality feature basis for subsequent user query matching, and improves the accuracy of the teaching assistant robot recommendation. In view of the problems that the existing teaching assistant robot processing system cannot automatically identify and highlight the real core knowledge points and is easily disturbed by noisy text, and lacks dynamic adaptation to the importance of information of different dimensions, the recommendation results of the teaching assistant robot are often mixed with marginal or irrelevant content, and it is difficult for users to quickly lock the required knowledge points, this solution generates the initial cluster center based on the word importance feature vector and initializes the dimension weights, constructs the membership matrix according to the dual-feature weighted distance, updates the cluster center according to the credibility index, combines the cluster contribution, part of speech influence, distribution concentration and variance to update the dimension weights, judges the convergence based on the rate of change of the objective function value and the maximum moving distance of the cluster center, obtains the clustering result, reduces the interference of noisy text on the clustering, makes the recommendation results of the teaching assistant robot stably focus on the core knowledge points, and allows users to quickly lock the required core information from the recommendation list.

[0004] The teaching assistant robot processing system based on natural language processing provided by the present invention includes a knowledge point text acquisition module, a knowledge point text preprocessing module, a knowledge point text vectorization processing module, a knowledge point text clustering module and a teaching assistant robot processing module;

[0005] The knowledge point text collection module is used by the teaching assistant robot to collect knowledge point texts;

[0006] The knowledge point text preprocessing module is used by the teaching assistant robot to preprocess the collected knowledge point texts to obtain a global vocabulary and a knowledge point text dataset;

[0007] The knowledge point text vectorization processing module calculates word frequency and inverse document frequency, introduces a chaotic adaptive wandering strategy to improve the dolphin swarm algorithm, obtains the part-of-speech influence, obtains the distribution concentration based on the distribution entropy, generates a word importance feature vector, and splices it with the semantic feature vector to obtain a knowledge point text vector;

[0008] The knowledge point text clustering module generates initial cluster centers based on word importance feature vectors and initializes dimension weights. It constructs a membership matrix based on the dual-feature weighted distance, updates cluster centers based on the credibility index, combines cluster contribution, part-of-speech influence, distribution concentration, and variance to update dimension weights, and determines convergence based on the rate of change of the objective function value and the maximum moving distance of the cluster center to obtain the clustering result.

[0009] The teaching assistant robot processing module receives the real-time text input by the user, selects the target cluster based on the clustering result, and generates a recommendation list.

[0010] Furthermore, the knowledge point text preprocessing module is a preprocessing module performed by the teaching assistant robot on the collected knowledge point texts. The preprocessing includes text cleaning, text segmentation, part-of-speech tagging and data set construction to obtain a global vocabulary and a knowledge point text data set.

[0011] Furthermore, the knowledge point text vectorization processing module performs vectorization processing on each knowledge point text in the knowledge point text dataset based on natural language processing technology. The knowledge point text vectorization processing module is provided with a word frequency and inverse document frequency calculation unit, a part of speech influence calculation unit, a distribution concentration calculation unit, a word importance feature value calculation unit, and a text vector generation unit; specifically, it includes the following contents:

[0012] Term frequency and inverse document frequency calculation unit: For each knowledge point text d in the knowledge point text dataset A, traverse each word t in the global vocabulary Z, count the number of times t appears in d and the number of times it appears in A, and count the total number of words in d and the total number of texts in A, and obtain the term frequency of t in d and the inverse document frequency in A;

[0013] Part-of-speech influence calculation unit: This unit uses the improved Dolphin Swarm algorithm to obtain the optimal part-of-speech influence for each type of part-of-speech, and then obtains the part-of-speech influence of each word in the global vocabulary. This unit includes the following:

[0014] Initial dolphin position; establish a search space for all part-of-speech categories in the global vocabulary, randomly initialize Q dolphin positions in the search space, use each dolphin position to represent a group of part-of-speech influence combinations, vectorize each knowledge point text in the knowledge point text dataset based on each group of part-of-speech influence combinations, and then use the K-Means algorithm to cluster all knowledge point text vectors. The F1 score of the clustering result is used as the fitness value of the corresponding dolphin position;

[0015] Dolphin position update; The dolphin group algorithm simulates the predation behavior of dolphins, including four key stages: searching, calling, receiving, and hunting. After the predation stage, a chaotic adaptive swimming strategy is added. When the fitness value of the global optimal position has not changed after five consecutive iterations, the chaotic adaptive swimming strategy is activated. The chaotic adaptive swimming strategy includes the following:

[0016] Generate a chaotic adaptive swimming path; sort all dolphin positions in the dolphin group in descending order of fitness value, select the top 20% dolphin positions as high-quality positions, pre-set the number of swimming steps to M, generate a direction factor based on the Logistic chaos value, and generate an adaptive step length based on the Euclidean distance between the high-quality position and the global optimal position. Combine the direction factor and the adaptive step length to obtain a chaotic adaptive swimming path corresponding to each high-quality position;

[0017] Update; set the dynamic boundary, map the chaotic adaptive wandering path into the dynamic boundary, generate a new position for each high-quality position, and if the fitness value of the new position is greater than the fitness value of the corresponding high-quality position, replace the corresponding high-quality position with the new position; otherwise, retain the original high-quality position;

[0018] The optimal part-of-speech influence is determined; when the fitness value of a dolphin position is greater than the pre-set fitness threshold γ, the part-of-speech influence combination represented by the dolphin position is taken as the optimal part-of-speech influence combination, and the part-of-speech influence of each word in the global vocabulary is obtained based on the optimal part-of-speech influence combination; otherwise, if the maximum number of iterations is reached, the number of iterations is reset to 0 and the dolphin position is reset; otherwise, the number of iterations is increased by 1 and the dolphin position update is continued;

[0019] Distribution concentration calculation unit: For each word t in the global vocabulary Z, count the number of knowledge point texts a containing t in the knowledge point text dataset A. If a knowledge point text d contains t, then the probability of d appearing in all knowledge point texts containing t is for ,otherwise ; Calculate the distribution entropy of t based on probability, normalize the distribution entropy, and obtain the distribution concentration of t;

[0020] Word importance feature value calculation unit: combines word frequency, inverse document frequency, part-of-speech influence and distribution concentration to obtain the word importance feature value of each word in the global vocabulary in each knowledge point text;

[0021] Text vector generation unit: for each knowledge point text d in the knowledge point text dataset A, arrange the word importance feature values ​​of all words in d according to the word order in the global vocabulary Z to form the word importance feature vector corresponding to d, and then generate the semantic feature vector of each d based on the BERT model, concatenate the word importance feature vector and the semantic feature vector to obtain the knowledge point text vector of d, and construct a knowledge point text vector dataset based on the knowledge point text vectors corresponding to all knowledge point texts in A.

[0022] Furthermore, the knowledge point text clustering module clusters all knowledge point text vectors in the knowledge point text vector dataset. The knowledge point text clustering module is provided with a clustering parameter initialization unit, a membership matrix construction unit, a cluster center update unit, a dimension weight update unit, and a convergence judgment unit; specifically, it includes the following contents:

[0023] Clustering parameter initialization unit: Calculate the sampling probability of each knowledge point text vector based on the word importance feature vector, perform weighted sampling G times from the knowledge point text vector dataset B according to the sampling probability, obtain G initial cluster centers, and set the weight of each dimension of the word importance feature vector and the semantic feature vector respectively;

[0024] Membership matrix construction unit; considering the weight of each dimension of the word importance feature vector and the semantic feature vector, calculating the dual-feature weighted distance between each knowledge point text vector and each current cluster center, assigning each knowledge point text vector to the cluster with the smallest dual-feature weighted distance, and obtaining the membership matrix;

[0025] Cluster center update unit; calculate the credibility index of each knowledge point text vector based on the distribution concentration, and calculate the center l of the word importance feature vector and semantic feature vector of all knowledge point text vectors in each cluster according to the credibility index wo and l se , will l wo and l se Splice to get the new cluster center of each cluster;

[0026] Dimension weight update unit: calculates the cluster contribution of each dimension in the word importance feature vector and the semantic feature vector respectively, then combines the part-of-speech influence and distribution concentration to update the weight of each dimension in the word importance feature vector, and updates the weight of each dimension in the semantic feature vector based on the variance of the dimension;

[0027] Convergence judgment unit; calculate the objective function value based on the dual-feature weighted distance and the credibility index. If the rate of change of the objective function value is less than or equal to 10 -5 And the maximum moving distance among all cluster centers is less than or equal to 10 -3 , it has converged, the clustering is terminated and the clustering result is output; otherwise, if the preset maximum number of updates is reached, the number of updates is reset to 0 and the process is transferred to the clustering parameter initialization unit; otherwise, the number of updates is increased by 1 and the process is transferred to the membership matrix construction unit.

[0028] Furthermore, the teaching assistant robot processing module is a module in which the teaching assistant robot receives real-time text input by the user, performs text cleaning, text segmentation, part-of-speech tagging and vectorization processing on the real-time text to obtain a real-time text vector, calculates the cosine similarity between the real-time text vector and each cluster center in the clustering result, selects the cluster to which the cluster center with the highest cosine similarity belongs as the target cluster, and then calculates the cosine similarity between the real-time text vector and the text vector of each knowledge point in the target cluster, arranges all corresponding knowledge point texts in the target cluster in descending order according to the size of the cosine similarity, and generates a recommendation list.

[0029] The beneficial effects achieved by the present invention using the above scheme are as follows:

[0030] (1) In view of the problem that the existing teaching assistant robot processing system only relies on word frequency-inverse document frequency or only relies on semantic vectors, which cannot accurately reflect the semantics and core information of knowledge points, resulting in difficulty in finding relevant knowledge points when users query and match, affecting the accuracy of teaching assistant robot recommendations, this solution calculates word frequency and inverse document frequency to quantify the local occurrence frequency and global rarity of words; introduces a chaotic adaptive wandering strategy to improve the dolphin group algorithm to obtain the influence of parts of speech, so that the feature vector can accurately reflect the difference in the contribution of parts of speech to the semantics of knowledge points; obtains the distribution concentration based on distribution entropy, so that the feature vector can highlight the core terms of concentrated distribution; generates word importance feature vectors, so that the vector can comprehensively characterize the contribution of words to knowledge points; and splices them with semantic feature vectors to obtain knowledge point text vectors. The generated vectors take into account both word importance and deep semantics, accurately characterize the core features of knowledge points, provide a high-quality feature basis for subsequent user query matching, and improve the accuracy of teaching assistant robot recommendations.

[0031] (2) In view of the problems in the existing teaching assistant robot processing system that it cannot automatically identify and highlight the real core knowledge points and is easily disturbed by noise text, and lacks dynamic adaptation to the importance of information of different dimensions, the recommendation results of the teaching assistant robot are often mixed with marginal or irrelevant content, making it difficult for users to quickly lock the required knowledge points. This solution generates the initial cluster center based on the word importance feature vector and initializes the dimension weight to make the initial center closer to the core knowledge points and reduce the interference of noise text on the initial clustering; constructs the membership matrix based on the dual-feature weighted distance, and considers the word importance feature and semantic feature at the same time, which can accurately divide the core and marginal content and reduce the mixing of irrelevant text into the core cluster; updates the cluster center according to the credibility index, making the cluster center more biased towards the characteristics of the core knowledge points; combines the cluster contribution, word part influence, distribution concentration and variance to update the dimension weight, so that the clustering process can dynamically focus on the most critical dimension for distinguishing the core knowledge points; judges the convergence based on the rate of change of the objective function value and the maximum moving distance of the cluster center to obtain the clustering result, so that the recommendation results of the teaching assistant robot can stably focus on the core knowledge points, and the user can quickly lock the required core information from the recommendation list. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 A schematic diagram of the teaching assistant robot processing system based on natural language processing provided by the present invention;

[0033] Figure 2 This is a schematic diagram of the knowledge point text vectorization processing module;

[0034] Figure 3 Schematic diagram of the knowledge point text clustering module.

[0035] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention. DETAILED DESCRIPTION

[0036] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments; based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0037] In the description of the present invention, it should be understood that terms such as "upper", "lower", "front", "back", "left", "right", "top", "bottom", "inside" and "outside" indicating directions or positional relationships are based on the directions or positional relationships shown in the accompanying drawings. They are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific direction, be constructed and operated in a specific direction. Therefore, they should not be understood as limiting the present invention.

[0038] Example 1, see Figure 1 The teaching assistant robot processing system based on natural language processing provided by the present invention includes a knowledge point text acquisition module, a knowledge point text preprocessing module, a knowledge point text vectorization processing module, a knowledge point text clustering module and a teaching assistant robot processing module;

[0039] The knowledge point text collection module is used by the teaching assistant robot to collect knowledge point text and send the data to the knowledge point text preprocessing module;

[0040] The knowledge point text preprocessing module receives data sent by the knowledge point text collection module, and the teaching assistant robot preprocesses the collected knowledge point text to obtain a global vocabulary and a knowledge point text dataset, and sends the data to the knowledge point text vectorization processing module;

[0041] The knowledge point text vectorization processing module receives data sent by the knowledge point text preprocessing module, calculates word frequency and inverse document frequency, introduces a chaotic adaptive wandering strategy to improve the dolphin group algorithm, obtains the part-of-speech influence, obtains the distribution concentration based on the distribution entropy, generates a word importance feature vector, and splices it with the semantic feature vector to obtain a knowledge point text vector, and sends the data to the knowledge point text clustering module;

[0042] The knowledge point text clustering module receives data sent by the knowledge point text vectorization processing module, generates initial cluster centers based on word importance feature vectors, initializes dimension weights, constructs a membership matrix based on dual-feature weighted distances, updates cluster centers based on credibility indexes, combines cluster contribution, part-of-speech influence, distribution concentration, and variance to update dimension weights, judges convergence based on the rate of change of the objective function value and the maximum moving distance of the cluster center, obtains clustering results, and sends the data to the teaching assistant robot processing module;

[0043] The teaching assistant robot processing module receives data sent by the knowledge point text clustering module, and the teaching assistant robot receives real-time text input by the user, selects a target cluster based on the clustering result, and generates a recommendation list.

[0044] Example 2, see Figure 1 This embodiment is based on the above embodiment. In the knowledge point text collection module, the teaching assistant robot collects knowledge point texts based on textbooks, courseware and exercise books.

[0045] Example 3, see Figure 1 This embodiment is based on the above embodiment. In the knowledge point text preprocessing module, the teaching assistant robot preprocesses the collected knowledge point text. The preprocessing includes text cleaning, text segmentation, part-of-speech tagging, and data set construction to obtain a global vocabulary and a knowledge point text data set.

[0046] The text cleaning is to remove special symbols, stop words and low-frequency words that appear less than 3 times in the knowledge point text;

[0047] The text segmentation is to use Jieba tool and NLTK tool to segment Chinese and English respectively, and split each knowledge point text into vocabulary sequences;

[0048] The part-of-speech tagging is to identify the grammatical attributes of words and mark the part of speech of each word in the knowledge point text;

[0049] The data set construction is to construct a knowledge point text data set based on all knowledge point texts after text cleaning, text segmentation and part-of-speech tagging, and to count the vocabulary of all knowledge point texts to generate a global vocabulary table.

[0050] Example 4, see Figure 1 and Figure 2 This embodiment is based on the above embodiment. In the knowledge point text vectorization processing module, each knowledge point text in the knowledge point text data set is vectorized based on natural language processing technology. The knowledge point text vectorization processing module is provided with a word frequency and inverse document frequency calculation unit, a part of speech influence calculation unit, a distribution concentration calculation unit, a word importance feature value calculation unit, and a text vector generation unit; specifically, it includes the following contents:

[0051] Word frequency and inverse document frequency calculation unit; the importance of a word cannot be distinguished by the number of times it appears in the text. The local frequency of a word in a single text is quantified by word frequency, and the global rarity of a word in all texts is quantified by inverse document frequency, which comprehensively reflects the importance of the word. For each knowledge point text d in the knowledge point text dataset A, each word t in the global vocabulary Z is traversed, and the number of times t appears in d and the number of times it appears in A are counted. The total number of words in d and the total number of texts in A are also counted to obtain the word frequency of t in d and the inverse document frequency in A. The formula used is as follows:

[0052] ;

[0053] ;

[0054] Where, d i is the i-th knowledge point text in A, tj is the jth word in Z, It is t j In d i The number of times it appears in i It is d i The total number of words in It is t j In d i The word frequency in A contains t j The number of texts, N A is the total number of texts in A, It is t j Inverse document frequency in A, i and j are the knowledge point text index and vocabulary index respectively;

[0055] Part-of-speech influence calculation unit: Different parts of speech have different importance in knowledge points, but it is difficult to subjectively determine their weights, and traditional optimization algorithms are prone to falling into local optimality. Through the improved Dolphin Swarm Algorithm, the F1 score of knowledge point clustering is used as the fitness level to optimize and obtain the optimal influence of each part of speech. The importance of part of speech is objectively quantified, and the semantic characteristics of knowledge points are adapted. The improved algorithm avoids local optimality, improves optimization accuracy, and ensures that the part-of-speech influence meets clustering requirements. The improved Dolphin Swarm Algorithm is used to obtain the optimal part-of-speech influence of each part of speech, and then the part-of-speech influence of each word in the global vocabulary. This includes the following content:

[0056] Initial dolphin position; establish a search space for all part-of-speech categories in the global vocabulary, and randomly initialize Q dolphin positions in the search space, where Q=50; use each dolphin position to represent a part-of-speech influence combination of a group of part-of-speech categories, and vectorize each knowledge point text in the knowledge point text dataset based on each part-of-speech influence combination. Then use the K-Means algorithm to cluster all knowledge point text vectors, and use the F1 score of the clustering result as the fitness value of the corresponding dolphin position;

[0057] Dolphin position update; The dolphin group algorithm simulates the hunting behavior of dolphins, including four key stages: searching, calling, receiving, and hunting. The dolphin position is iteratively updated to find the optimal position. After the hunting stage, a chaotic adaptive swimming strategy is added. When the fitness value of the global optimal position does not change after five consecutive iterative updates, the chaotic adaptive swimming strategy is activated. The chaotic adaptive swimming strategy includes the following:

[0058] Generate a chaotic adaptive swimming path; sort all dolphin positions in the dolphin group in descending order according to fitness value, select the top 20% dolphin positions as high-quality positions, pre-set the number of swimming steps to M, generate a direction factor based on the Logistic chaos value, and generate an adaptive step length based on the Euclidean distance between the high-quality position and the global optimal position. Combine the direction factor and the adaptive step length to obtain the chaotic adaptive swimming path corresponding to each high-quality position; the formula used is as follows:

[0059] ;

[0060] ;

[0061] ;

[0062] Where, and They are the direction factor and adaptive step size of the qth high-quality position walking in the mth step at the pth iteration, z m is the Logistic chaos value corresponding to the m-th step walk, , M=10, is the Euclidean distance between the qth best position and the global optimal position at the pth iteration. The global optimal position is the dolphin position with the largest fitness value. is the maximum Euclidean distance between the dolphin position in the dolphin group and the global optimal position at the pth iteration, q, m and p are the high-quality position index, the walking step index and the iteration number index respectively. is the chaotic adaptive walking path of the qth high-quality position at the pth iteration;

[0063] Update; set the dynamic boundary, map the chaotic adaptive wandering path into the dynamic boundary, generate a new position for each high-quality position, and if the fitness value of the new position is greater than the fitness value of the corresponding high-quality position, replace the corresponding high-quality position with the new position; otherwise, retain the original high-quality position; the formula used is as follows:

[0064] ;

[0065] ;

[0066] ;

[0067] ;

[0068] Where, and are the upper and lower boundaries of the wth dimension at the pth iteration, and They are the upper and lower boundaries of the search space in the wth dimension, p max is the maximum number of iterations, p max =100, and are the values ​​of the qth quality position and the global optimal position in the wth dimension at the pth iteration, yes The corresponding new position, and are the adaptation adjustment coefficient and fitness value of the qth high-quality position at the pth iteration, is the average fitness value of all dolphin positions in the dolphin group at the pth iteration, L min and L max are the minimum and maximum values ​​in all chaotic adaptive walk paths, respectively, and w is the dimension index;

[0069] The optimal part-of-speech influence is determined. When the fitness value of a dolphin position is greater than the pre-set fitness threshold γ, γ = 0.85, the part-of-speech influence combination represented by the dolphin position is taken as the optimal part-of-speech influence combination, and the part-of-speech influence of each word in the global vocabulary is obtained based on the optimal part-of-speech influence combination. Otherwise, if the maximum number of iterations is reached, the number of iterations is reset to 0 and the dolphin position is reset. Otherwise, the number of iterations is increased by 1 and the dolphin position update is continued.

[0070] Distribution concentration calculation unit; the distribution range of vocabulary in the knowledge point text affects its representativeness for a specific knowledge point. Through distribution entropy and normalization calculation, the distribution concentration of vocabulary in the knowledge point text is quantified, reflecting the close relationship between vocabulary and specific knowledge points. Concentrated vocabulary is more likely to be the core term of the knowledge point, providing a supplementary dimension for word importance assessment; for each vocabulary t in the global vocabulary table Z, the number of knowledge point texts containing t in the knowledge point text dataset A is counted. If a knowledge point text d contains t, then the probability of d appearing in all knowledge point texts containing t is for ,otherwise ; Calculate the distribution entropy of t based on probability, normalize the distribution entropy, and obtain the distribution concentration of t; the formula used is as follows:

[0071] ;

[0072] ;

[0073] Where, and They are t j The distribution entropy and distribution concentration of Is included in t j All knowledge points in the texti The probability of occurrence, a j A contains t j The number of knowledge point texts;

[0074] Word importance feature value calculation unit; a single dimension cannot fully reflect the importance of a word in a knowledge point. TF, IDF, part-of-speech influence, and distribution concentration are integrated to obtain the comprehensive importance feature value of a word in a specific text. Multi-dimensional feature fusion more comprehensively depicts the contribution of a word to a knowledge point, avoids the one-sidedness of a single indicator, and improves the accuracy of feature representation; word frequency, inverse document frequency, part-of-speech influence, and distribution concentration are integrated to obtain the word importance feature value of each word in the global vocabulary in each knowledge point text; the formula used is as follows:

[0075] ;

[0076] Where, It is t j In d i The word importance feature value in , It is t j The influence of part of speech;

[0077] Text vector generation unit; using only word importance feature vectors cannot capture the deep semantics of the text, and using only BERT semantic vectors cannot reflect the specific importance of vocabulary. The word importance feature vectors and BERT semantic feature vectors are concatenated to form a knowledge point text vector that takes into account both vocabulary importance and deep semantics. This not only retains the importance information at the vocabulary level, but also captures the overall semantics of the text, making the vector representation more comprehensive and laying a high-quality feature foundation for subsequent clustering and matching; for each knowledge point text d in the knowledge point text dataset A, the word importance feature values ​​of all words in d are arranged according to the vocabulary order in the global vocabulary table Z to form a J-dimensional word importance feature vector corresponding to d, and then a semantic feature vector of each d is generated based on the BERT model. The word importance feature vector and the semantic feature vector are concatenated to obtain the knowledge point text vector of d, and a knowledge point text vector dataset is constructed based on the knowledge point text vectors corresponding to all knowledge point texts in A; the formula used is as follows:

[0078] ;

[0079] Where, It is d i The word importance feature vector of Z, J is the number of words in Z, 、 and They are The values ​​in the 1st, jth, and Jth dimensions; if The vocabulary t corresponding to the j-th dimension in j In di If it appears in ,otherwise .

[0080] By performing the above operations, the existing teaching assistant robot processing system only relies on word frequency-inverse document frequency or only relies on semantic vectors, which cannot accurately reflect the semantics and core information of knowledge points, resulting in difficulty in finding relevant knowledge points when users query and match, affecting the recommendation accuracy of the teaching assistant robot. This solution calculates word frequency and inverse document frequency to quantify the local occurrence frequency and global rarity of words; introduces a chaotic adaptive wandering strategy to improve the dolphin swarm algorithm to obtain the part-of-speech influence, so that the feature vector can accurately reflect the difference in the contribution of part-of-speech to the semantics of knowledge points; obtains the distribution concentration based on the distribution entropy, so that the feature vector can highlight the core terms of the concentrated distribution; generates a word importance feature vector, so that the vector can comprehensively characterize the contribution of the vocabulary to the knowledge point; and splices it with the semantic feature vector to obtain the knowledge point text vector. The generated vector takes into account both word importance and deep semantics, accurately characterizes the core features of the knowledge point, provides a high-quality feature foundation for subsequent user query matching, and improves the accuracy of the teaching assistant robot recommendation.

[0081] Example 5, see Figure 1 and Figure 3 This embodiment is based on the above embodiment. In the knowledge point text clustering module, all knowledge point text vectors in the knowledge point text vector dataset are clustered. The knowledge point text clustering module is provided with a clustering parameter initialization unit, a membership matrix construction unit, a cluster center update unit, a dimension weight update unit, and a convergence judgment unit; specifically, it includes the following contents:

[0082] Clustering parameter initialization unit; Randomly initializing the cluster center may cause the clustering results to deviate from the actual knowledge point distribution, and unreasonable setting of the dimension weights of word importance and semantic features will affect the clustering direction. The initial cluster center is selected based on the sampling probability weight of the word importance feature vector, and the dimension weights are initialized reasonably to make the initial center closer to the actual knowledge point cluster center. The initial weight value adapts to the feature importance, reduces the number of subsequent iterations, and improves the quality of the clustering starting point; the sampling probability of each knowledge point text vector is calculated based on the word importance feature vector, and the number of cluster categories G is determined based on the silhouette coefficient. According to the sampling probability, weighted sampling G times is performed from the knowledge point text vector dataset B to obtain G initial cluster centers, and the weights of each dimension of the word importance feature vector and the semantic feature vector are set respectively; the formula used is as follows:

[0083] ;

[0084] ;

[0085] ;

[0086] Where C v Yes v The sampling probability of and They are o v and o k The value of the corresponding word importance feature vector in the jth dimension, o v and o k are the vth and kth knowledge point text vectors in B, respectively. v and k are the knowledge point text vector indices. is the weight of the j-th dimension of the word importance feature vector, is the weight of the rth dimension of the semantic feature vector, is the maximum value of all knowledge point text vectors in B in the jth dimension, and R is the dimension of the semantic feature vector;

[0087] Membership matrix construction unit; Calculating distance based only on word importance or semantic single feature may lead to clustering bias. Calculate the dual-feature weighted distance of word importance feature and semantic feature, taking into account both lexical importance and semantic similarity, so that the clustering results are more consistent with the actual association of knowledge points; Consider the weight of each dimension of word importance feature vector and semantic feature vector at the same time, calculate the dual-feature weighted distance from each knowledge point text vector to each current cluster center, assign each knowledge point text vector to the cluster to which the dual-feature weighted distance belongs to, and obtain the membership matrix ; The formula used is as follows:

[0088] ;

[0089] in, is the g-th cluster center at the n-th update, Yes v and The double-feature weighted distance between is the weight of the j-th dimension of the word importance feature vector at the n-th update, is the weight of the rth dimension of the semantic feature vector at the nth update, yes The value of the corresponding word importance feature vector in the jth dimension, and They are o v and The value of the corresponding semantic feature vector in the rth dimension, is the membership matrix at the nth update, yes The value in row v, column g; when o v Assigned to When it belongs to a cluster, ,otherwise, ;

[0090] Cluster center update unit; The credibility of different knowledge point texts is different. Direct averaging will cause the cluster center to deviate from the true core. The text credibility index is calculated based on the distribution concentration, and the cluster center of the word importance and semantic features is weighted to update, so that the cluster center is more biased towards the knowledge point text with high credibility, improving the representativeness of the center and making the text in the cluster more consistent; The credibility index of each knowledge point text vector is calculated based on the distribution concentration. According to the credibility index, the center l of the word importance feature vector and the semantic feature vector of all knowledge point text vectors in each cluster is calculated respectively. wo and l se , will l wo and l se Splice to get the new cluster center of each cluster; the formula used is as follows:

[0091] ;

[0092] ;

[0093] ;

[0094] in, Yes v The credibility index, d v Yes v The corresponding knowledge point text, is an indicator function; if ,but ,otherwise, ; is the central value of the word importance feature vector of all knowledge point text vectors in the g-th cluster in the j-th dimension during the n+1-th update. is the central value of the semantic feature vector of all knowledge point text vectors in the g-th cluster in the r-th dimension during the n+1-th update;

[0095] Dimension weight update unit; the role of each dimension of word importance and semantic features in the clustering process changes with iteration, and fixed weights cannot adapt to dynamic needs. Based on cluster contribution, combined with word influence, distribution concentration and variance, the weight is dynamically updated to highlight the dimensions that are more important for cluster differentiation, suppress irrelevant dimensions, adapt the weight to the clustering process, and improve clustering accuracy; calculate the cluster contribution of each dimension in the word importance feature vector and the semantic feature vector respectively, then combine the word influence and distribution concentration to update the weight of each dimension in the word importance feature vector, and update the weight of each dimension in the semantic feature vector based on the variance of the dimension; the formula used is as follows:

[0096] ;

[0097] ;

[0098] ;

[0099] ;

[0100] in, is the clustering contribution of the jth dimension of the word importance feature vector at the nth update, is the clustering contribution of the rth dimension of the semantic feature vector at the nth update, is the weight of the j-th dimension of the word importance feature vector at the n+1th update, is the weight of the rth dimension of the semantic feature vector at the n+1th update, δ and φ are the word sparsity penalty coefficient and semantic sparsity penalty coefficient, δ=0.1, φ=0.05, is the variance of the semantic feature vector of all knowledge point text vectors in B in the rth dimension;

[0101] Convergence judgment unit; cluster update needs to determine the termination condition to avoid insufficient or excessive updates, and judge whether it is converged by the rate of change of the objective function and the maximum moving distance of the cluster center, control the termination of the update, ensure that the clustering results are stable and optimal, balance the processing efficiency and result quality, and avoid invalid updates; calculate the objective function value based on the dual-feature weighted distance and the credibility index. If the rate of change of the objective function value is less than or equal to 10 -5 And the maximum moving distance among all cluster centers is less than or equal to 10 -3 , it has converged, the clustering is terminated and the clustering results are output; otherwise, if the preset maximum number of updates n is reached max , the update number is reset to 0 and the process is transferred to the clustering parameter initialization unit; otherwise, the update number is increased by 1 and the process is transferred to the membership matrix construction unit. The formula used is as follows:

[0102] ;

[0103] ;

[0104] ;

[0105] Where, and are the objective function values ​​at the nth and n-1th updates, respectively. is the rate of change of the objective function value at the nth update, is the gth cluster center at the n-1th update, n max =50, is the maximum moving distance among all cluster centers at the nth update, is the L2 paradigm, is the smoothing term, .

[0106] By performing the above operations, we address the problems in existing teaching assistant robot processing systems, which often fail to automatically identify and highlight core knowledge points, are easily distracted by noisy text, and lack dynamic adaptation to the importance of information in different dimensions. This results in the teaching assistant robot's recommendation results often being mixed with marginal or irrelevant content, making it difficult for users to quickly identify the desired knowledge points. This solution generates initial cluster centers based on word importance feature vectors and initializes dimension weights to align the initial centers more closely with core knowledge points, reducing the interference of noisy text on the initial clusters. A membership matrix is ​​constructed based on dual-feature weighted distance, taking into account both word importance and semantic features. This accurately separates core and marginal content and reduces the incorporation of irrelevant text into core clusters. Cluster centers are updated based on the credibility index, aligning them more closely with the characteristics of core knowledge points. Dimension weights are updated by combining cluster contribution, part-of-speech influence, distribution concentration, and variance, enabling the clustering process to dynamically focus on the dimensions most critical for distinguishing core knowledge points. Convergence is determined based on the rate of change of the objective function value and the maximum moving distance of the cluster centers to obtain clustering results. This ensures that the teaching assistant robot's recommendations are stably focused on core knowledge points, allowing users to quickly identify the desired core information from the recommendation list.

[0107] Example 6, see Figure 1 This embodiment is based on the above embodiment. In the teaching assistant robot processing module, the real-time text input by the user needs to be quickly matched to the most relevant knowledge point, otherwise effective assistance cannot be provided in time. After the real-time text is vectorized, the most relevant cluster clusters and knowledge points within the clusters are matched by cosine similarity to generate a recommendation list. Based on the clustering results and similarity calculation, the recommended knowledge points are ensured to be highly relevant to the user input. Sorting by similarity improves the efficiency of the user in obtaining information and quickly meets learning needs. The teaching assistant robot receives the real-time text input by the user, performs text cleaning, text segmentation, part-of-speech tagging and vectorization processing on the real-time text to obtain a real-time text vector, calculates the cosine similarity between the real-time text vector and each cluster center in the clustering result, selects the cluster to which the cluster center with the highest cosine similarity belongs as the target cluster, and then calculates the cosine similarity between the real-time text vector and the text vector of each knowledge point in the target cluster. All corresponding knowledge point texts in the target cluster are arranged in descending order according to the size of the cosine similarity to generate a recommendation list.

[0108] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0109] While the embodiments of the present invention have been shown and described, it will be apparent to those skilled in the art that various changes, modifications, substitutions, and alterations can be made to the embodiments without departing from the principles and spirit of the invention.

[0110] The present invention and its embodiments are described above. This description is not restrictive. The drawings show only one embodiment of the present invention, and the actual structure is not limited thereto. In short, if a person skilled in the art is inspired by this and, without departing from the purpose of the present invention, designs structures and embodiments similar to this technical solution without inventiveness, they shall fall within the scope of protection of the present invention.

Claims

1. A teaching assistant robot processing system based on natural language processing, characterized by: It includes knowledge point text collection module, knowledge point text preprocessing module, knowledge point text vectorization processing module, knowledge point text clustering module and teaching assistant robot processing module; The knowledge point text collection module is used by the teaching assistant robot to collect knowledge point texts; The knowledge point text preprocessing module is used by the teaching assistant robot to preprocess the collected knowledge point texts to obtain a global vocabulary and a knowledge point text dataset; The knowledge point text vectorization processing module calculates word frequency and inverse document frequency, introduces a chaotic adaptive wandering strategy to improve the dolphin swarm algorithm, obtains the part-of-speech influence, obtains the distribution concentration based on the distribution entropy, generates a word importance feature vector, and splices it with the semantic feature vector to obtain a knowledge point text vector; The knowledge point text clustering module generates initial cluster centers based on word importance feature vectors and initializes dimension weights. It constructs a membership matrix based on the dual-feature weighted distance, updates cluster centers based on the credibility index, combines cluster contribution, part-of-speech influence, distribution concentration, and variance to update dimension weights, and determines convergence based on the rate of change of the objective function value and the maximum moving distance of the cluster center to obtain the clustering result. The teaching assistant robot processing module receives the real-time text input by the user, selects the target cluster based on the clustering result, and generates a recommendation list.

2. The teaching assistant robot processing system based on natural language processing according to claim 1, characterized in that: The knowledge point text vectorization processing module is based on natural language processing technology to perform vectorization processing on each knowledge point text in the knowledge point text data set. The knowledge point text vectorization processing module is provided with a word frequency and inverse document frequency calculation unit, a part of speech influence calculation unit, a distribution concentration calculation unit, a word importance feature value calculation unit and a text vector generation unit; specifically, Includes the following: Term frequency and inverse document frequency calculation unit: For each knowledge point text d in the knowledge point text dataset A, traverse each word t in the global vocabulary Z, count the number of times t appears in d and the number of times it appears in A, and count the total number of words in d and the total number of texts in A, and obtain the term frequency of t in d and the inverse document frequency in A; Part of speech influence calculation unit; By improving the Dolphin Swarm Algorithm, we can obtain the optimal part-of-speech influence of each part-of-speech category, and then obtain the part-of-speech influence of each word in the global vocabulary. Distribution concentration calculation unit; Word importance feature value calculation unit: combines word frequency, inverse document frequency, part-of-speech influence and distribution concentration to obtain the word importance feature value of each word in the global vocabulary in each knowledge point text; Text vector generation unit: for each knowledge point text d in the knowledge point text dataset A, arrange the word importance feature values ​​of all words in d according to the word order in the global vocabulary Z to form the word importance feature vector corresponding to d, and then generate the semantic feature vector of each d based on the BERT model, concatenate the word importance feature vector and the semantic feature vector to obtain the knowledge point text vector of d, and construct a knowledge point text vector dataset based on the knowledge point text vectors corresponding to all knowledge point texts in A.

3. The teaching assistant robot processing system based on natural language processing according to claim 2, characterized in that: The part-of-speech influence calculation unit specifically includes the following contents: Initial dolphin position; establish a search space for all part-of-speech categories in the global vocabulary, randomly initialize Q dolphin positions in the search space, use each dolphin position to represent a group of part-of-speech influence combinations, vectorize each knowledge point text in the knowledge point text dataset based on each group of part-of-speech influence combinations, and then use the K-Means algorithm to cluster all knowledge point text vectors. The F1 score of the clustering result is used as the fitness value of the corresponding dolphin position; Dolphin position update; The dolphin group algorithm simulates the predation behavior of dolphins, including four key stages: searching, calling, receiving, and hunting. After the predation stage, a chaotic adaptive swimming strategy is added. When the fitness value of the global optimal position has not changed after five consecutive iterations, the chaotic adaptive swimming strategy is activated. The chaotic adaptive swimming strategy includes the following: Generate chaotic adaptive wandering paths; Update; set the dynamic boundary, map the chaotic adaptive wandering path into the dynamic boundary, generate a new position for each high-quality position, and if the fitness value of the new position is greater than the fitness value of the corresponding high-quality position, replace the corresponding high-quality position with the new position; otherwise, retain the original high-quality position; The optimal part-of-speech influence is determined; when the fitness value of a dolphin position is greater than a preset fitness threshold, the part-of-speech influence combination represented by the dolphin position is taken as the optimal part-of-speech influence combination, and the part-of-speech influence of each word in the global vocabulary is obtained based on the optimal part-of-speech influence combination; otherwise, if the maximum number of iterations is reached, the number of iterations is reset to 0 and the dolphin position is re-initialized; otherwise, the number of iterations is increased by 1 and the dolphin position update is continued.

4. The teaching assistant robot processing system based on natural language processing according to claim 3 is characterized in that: The chaotic adaptive swimming path is generated by arranging all dolphin positions in the dolphin group in descending order according to fitness values, screening the top 20% dolphin positions as high-quality positions, presetting the number of swimming steps to M, generating a direction factor based on the Logistic chaos value, and generating an adaptive step length based on the Euclidean distance between the high-quality positions and the global optimal position. The direction factor and the adaptive step length are combined to obtain a chaotic adaptive swimming path corresponding to each high-quality position.

5. The teaching assistant robot processing system based on natural language processing according to claim 2, characterized in that: The distribution concentration calculation unit is to count the number a of knowledge point texts containing t in the knowledge point text data set A for each word t in the global vocabulary table Z. If a knowledge point text d contains t, then the probability of d appearing in all knowledge point texts containing t is for ,otherwise ; The distribution entropy of t is calculated based on the probability, and the distribution entropy is normalized to obtain the distribution concentration of t.

6. The teaching assistant robot processing system based on natural language processing according to claim 1, characterized in that: The knowledge point text clustering module clusters all the knowledge point text vectors in the knowledge point text vector data set. The knowledge point text clustering module is provided with a clustering parameter initialization unit, a membership matrix construction unit, a cluster center update unit, a dimension weight update unit and a convergence judgment unit. Includes the following: Clustering parameter initialization unit; Calculate the sampling probability of each knowledge point text vector based on the word importance feature vector. Perform weighted sampling G times from the knowledge point text vector dataset B according to the sampling probability to obtain G initial cluster centers. Set the weight of each dimension of the word importance feature vector and the semantic feature vector respectively. Membership matrix construction unit; Cluster center update unit; The credibility index of each knowledge point text vector is calculated based on the distribution concentration. According to the credibility index, the center l of the word importance feature vector and semantic feature vector of all knowledge point text vectors in each cluster is calculated respectively. wo and l se , will l wo and l se Splice to get the new cluster center of each cluster; Dimension weight update unit; Convergence judgment unit; calculate the objective function value based on the dual-feature weighted distance and the credibility index. If the rate of change of the objective function value is less than or equal to 10 -5 And the maximum moving distance among all cluster centers is less than or equal to 10 -3 , it has converged, the clustering is ended and the clustering results are output; Otherwise, if the preset maximum number of updates is reached, the number of updates is reset to 0 and the process goes to the clustering parameter initialization unit; Otherwise, increase the update count by 1 and go to the membership matrix construction unit.

7. The teaching assistant robot processing system based on natural language processing according to claim 6, characterized in that: The membership matrix construction unit simultaneously considers the weight of each dimension of the word importance feature vector and the semantic feature vector, calculates the dual-feature weighted distance from each knowledge point text vector to each current cluster center, and assigns each knowledge point text vector to the cluster to which the cluster center with the smallest dual-feature weighted distance belongs to obtain the membership matrix.

8. The teaching assistant robot processing system based on natural language processing according to claim 6, characterized in that: The dimension weight updating unit calculates the clustering contribution of each dimension in the word importance feature vector and the semantic feature vector respectively, then combines the part-of-speech influence and distribution concentration to update the weight of each dimension in the word importance feature vector, and updates the weight of each dimension in the semantic feature vector based on the variance of the dimension.

9. The teaching assistant robot processing system based on natural language processing according to claim 1, characterized in that: The knowledge point text preprocessing module is used by the teaching assistant robot to preprocess the collected knowledge point text. The preprocessing includes text cleaning, text segmentation, part-of-speech tagging and data set construction to obtain a global vocabulary and a knowledge point text data set.

10. The teaching assistant robot processing system based on natural language processing according to claim 1, characterized in that: The teaching assistant robot processing module is a module in which the teaching assistant robot receives real-time text input by the user, performs text cleaning, text segmentation, part-of-speech tagging and vectorization on the real-time text to obtain a real-time text vector, calculates the cosine similarity between the real-time text vector and each cluster center in the clustering result, selects the cluster to which the cluster center with the highest cosine similarity belongs as the target cluster, then calculates the cosine similarity between the real-time text vector and the text vector of each knowledge point in the target cluster, arranges all corresponding knowledge point texts in the target cluster in descending order according to the size of the cosine similarity, and generates a recommendation list.

Images