Oral talent resource recommendation method, system and device and medium

Abstract

The invention provides an oral talent resource recommendation method, system and device and a medium, and the method comprises the steps: obtaining self-evaluation answers made by a user according to preset questions, carrying out the preliminary classification scoring of the self-evaluation answers through the preset weight of each preset question, and obtaining a first score of each designated oral talent dimension; obtaining a speech video uploaded by the user, performing audio feature analysis on the speech video to generate a feature vector sequence, and calculating a second score of each specified talent dimension based on the feature vector sequence; analyzing the first score and the second score corresponding to each specified talent dimension based on a multiple regression model to obtain a comprehensive score; and screening the oral resource of the corresponding authority according to the comprehensive score, and pushing the oral resource, so that the pushed oral resource is matched with the oral level of the user, and the pushing accuracy is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of speech analysis, and in particular to a method, system, device and medium for recommending eloquence resources. Background Art

[0002] At present, the system pushes eloquence training resources to users by matching the keywords entered by users with the keywords of resource titles, or by comparing the similarity between the keywords entered and the content of resources, and pushes all eloquence training resources with successful keyword matching or high similarity to users. Users can choose the required courses and resources to watch and learn according to their own needs. However, this system push method is relatively simple, and the system cannot push eloquence resources suitable for users to watch according to their current level, resulting in low matching between the resources pushed by the system and users, and poor push accuracy. Summary of the invention

[0003] The embodiment of the present invention provides a method, system, device and medium for recommending eloquence resources to solve the problems existing in the related technologies. The technical solution is as follows:

[0004] In a first aspect, an embodiment of the present invention provides a method for recommending eloquence resources, comprising:

[0005] Obtain the self-evaluation answers made by the user according to the preset questions, and preliminarily classify and score the self-evaluation answers according to the preset weight of each preset question to obtain the first score of each designated eloquence dimension;

[0006] Obtain the speech video uploaded by the user, perform audio feature analysis on the speech video to generate a feature vector sequence, and calculate the second score of each specified eloquence dimension based on the feature vector sequence;

[0007] The first and second scores corresponding to each designated eloquence dimension were analyzed based on a multiple regression model to obtain a comprehensive score;

[0008] According to the comprehensive score, select the eloquence resources with corresponding permissions and push them.

[0009] In one embodiment, the preset questions include questions corresponding to a plurality of specified eloquence dimensions, and the specified eloquence dimensions include fluency, pronunciation accuracy, speaking speed, vocabulary richness, sentence complexity, expressiveness, and content relevance.

[0010] In one embodiment, performing audio feature analysis on a speech video to generate a feature vector sequence includes:

[0011] The audio data corresponding to the speech video is divided into frames of fixed length, and the speech features of each frame are extracted using Mel-frequency cepstral coefficients to form a feature vector sequence.

[0012] In one embodiment, calculating the second score of each designated eloquence dimension based on the feature vector sequence includes:

[0013] The feature vector sequence is analyzed based on a pre-built deep neural network to obtain a second score for each specified eloquence dimension.

[0014] In one embodiment, it further includes:

[0015] According to the speech features of each frame in the speech video, the total pause duration and the average pause duration corresponding to each pause are counted;

[0016] The difference between the total pause duration and the average pause duration is calculated, and the rate of change of fluency over time is calculated based on the difference and a preset adjustment coefficient, and the fluency score is determined based on the rate of change.

[0017] In one embodiment, it further includes:

[0018] Extracting first-dimension features and second-dimension features related to expressiveness based on speech videos;

[0019] Mapping the first dimension feature and the second dimension feature to the real part and the imaginary part of the complex number respectively;

[0020] The expressiveness score is determined by analyzing the modulus and phase of the complex number.

[0021] In one embodiment, a method for calculating a comprehensive score includes:

[0022] The first score corresponding to each designated eloquence dimension and the second score corresponding to each designated eloquence dimension are used as the independent variables of the multiple regression model, and the comprehensive score is used as the dependent variable of the multiple regression model;

[0023] Determine the regression coefficient of each variable on the comprehensive score, and use the regression coefficient to perform multiple regression analysis on the first score corresponding to each specified eloquence dimension and the second score corresponding to each specified eloquence dimension to obtain the comprehensive score.

[0024] In a second aspect, an embodiment of the present invention provides an eloquence resource recommendation system, which executes the eloquence resource recommendation method as described above.

[0025] In a third aspect, an embodiment of the present invention provides an electronic device, the device comprising: a memory and a processor. The memory and the processor communicate with each other through an internal connection path, the memory is used to store instructions, the processor is used to execute the instructions stored in the memory, and when the processor executes the instructions stored in the memory, the processor executes the method in any one of the above-mentioned embodiments.

[0026] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium, which stores a computer program. When the computer program runs on a computer, the method in any one of the above-mentioned embodiments is executed.

[0027] The advantages or beneficial effects of the above technical solution include at least:

[0028] The present invention calculates a preliminary eloquence score through the user's answers to a self-assessment questionnaire on multiple eloquence dimensions, then scores each eloquence dimension according to the speech video uploaded by the user, uses the questionnaire question score and the video analysis result as independent variables, obtains a comprehensive score through a multivariate regression model, and finally pushes eloquence resources of corresponding authority to the user according to the comprehensive score, so that the pushed eloquence resources match the user's own eloquence level, thereby improving the push accuracy.

[0029] The above summary is for illustrative purposes only and is not intended to be limiting in any way. In addition to the illustrative aspects, embodiments and features described above, further aspects, embodiments and features of the present invention will be readily apparent by reference to the accompanying drawings and the following detailed description. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] In the accompanying drawings, unless otherwise specified, the same reference numerals throughout the multiple drawings represent the same or similar parts or elements. These drawings are not necessarily drawn to scale. It should be understood that these drawings only depict some embodiments disclosed in the present invention and should not be regarded as limiting the scope of the present invention.

[0031] Figure 1 A schematic diagram of a flow chart of a method for recommending eloquence resources according to the present invention;

[0032] Figure 2 The structure block diagram of an electronic device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0033] In the following, only some exemplary embodiments are briefly described. As those skilled in the art will appreciate, the described embodiments may be modified in various ways without departing from the spirit or scope of the present invention. Therefore, the drawings and descriptions are considered to be exemplary and non-restrictive in nature.

[0034] This embodiment provides a method for recommending eloquence resources, through which the matching degree between system push and user demand can be improved, and the push accuracy can be improved.

[0035] This embodiment can provide an interactive page for inputting basic user information, including name, email address, phone number, occupation, educational background, etc. Since the eloquence resource recommendation method of this embodiment can push corresponding resources according to the user's own eloquence level, after the user enters his own information in the information input interface, an account is created for the user, a unique user ID is generated for the account, and a corresponding user record is created in the database, so that the user can search for usage records and store the eloquence resources he needs.

[0036] The interactive page is provided with a self-assessment entrance and a video analysis entrance. Users can complete the self-assessment operation and / or the video upload operation according to their own needs, and analyze the self-assessment answers obtained from the self-assessment operation and / or the uploaded speech video to obtain eloquence resources that meet the user's own eloquence level.

[0037] When the user has only completed the self-assessment operation, the corresponding eloquence resource recommendation method is as follows:

[0038] The system obtains the user's self-assessment answers based on the preset questions, and makes a preliminary classification and scoring of the self-assessment answers based on the preset weights of each preset question, so as to assign the user to different eloquence level categories. Based on the user's preliminary classification score, the system can recommend courses and resources suitable for the user's current level category, thereby improving the match between the pushed resources and the user's level.

[0039] Among them, the preset questions include questions corresponding to multiple specified eloquence dimensions. Multiple specified eloquence dimensions can be pre-set according to actual needs. The specified eloquence dimensions include fluency, pronunciation accuracy, speaking speed, vocabulary richness, sentence complexity, expressiveness, and content relevance, etc. Set the corresponding self-assessment questions for each specified eloquence dimension, and display the self-assessment questions on the interactive page for users to answer.

[0040] Each preset question can be pre-configured with its corresponding preset weight. The score corresponding to each question can be obtained based on the self-assessment answer made by the user. The user's preliminary eloquence score can be calculated based on the score of each question and the weight corresponding to each question. The user's eloquence level can be classified into corresponding level grades based on the preliminary eloquence score. The authority corresponding to the level grade can be obtained based on the user's level grade. The system can then obtain the eloquence resources corresponding to the authority corresponding to the user's level grade and push them to the user, so that the pushed resources match the user's eloquence level grade, thereby improving the push accuracy.

[0041] For example, during the registration process on the interactive page, the user answered a self-assessment question about pronunciation accuracy, such as "How accurate do you think your pronunciation is in a speech? (1-5 points)", and the user chose 4 points. Assuming that the weight of this question is 0.3, the answer will be multiplied by the weight during the preliminary classification calculation, and the score for this question will be 1.2. After the questions corresponding to each eloquence dimension are answered, the scores of all questions are added up, and the average score is calculated to form the user's preliminary eloquence score. If the user's total preliminary score is 3.8, according to the predefined standards, this user may be classified as an intermediate eloquence level. Through such preliminary classification, the system can recommend courses and resources suitable for intermediate eloquence levels to users, and provide personalized learning paths and improvement suggestions.

[0042] In some embodiments, self-assessment questions can also be asked in the form of subjective questions. For example, the self-assessment question about pronunciation accuracy is "Please introduce the cultural characteristics of a certain place". Users can give free answers based on the self-assessment question. The system records the voice of the user when answering the question, analyzes the pronunciation accuracy of the voice, and obtains the corresponding score. During this period, the text of the user's answer can be obtained by voice-to-text conversion, and the standard pronunciation corresponding to each word or sentence in the text is compared with the user's voice, and the corresponding pronunciation accuracy score is obtained according to the similarity.

[0043] Among them, the formula for the preliminary eloquence score is:

[0044]

[0045] Among them, Si is the preliminary eloquence score of user i; N is the number of questionnaire questions; Wj is the weight of the jth question; Rij is the score of user i on the jth question.

[0046] The weights are adjusted by gradient descent method, and the formula is:

[0047]

[0048] Where: W' j is the adjusted weight; α is the learning rate; Aj is the average score of all users for the jth question, The average score of all questions.

[0049] Since the present embodiment optimizes the weights by the gradient descent method, the learning rate α can be initially set to 0.01 by the rule of thumb, and then dynamically adjusted during the training process, that is, as the training progresses, the learning rate gradually decreases, and its learning rate decay strategy can be that the learning rate is halved every 100 iterations; through the appropriate learning rate and its adjustment strategy, the efficiency and performance of the model training can be effectively improved, and the accuracy and reliability of the evaluation results can be ensured.

[0050] In some embodiments, the user may only complete the video upload operation, and the system may only analyze the speech video uploaded by the user to obtain the corresponding score, so as to make subsequent personalized recommendations and services based on the score.

[0051] Users can record speech videos on the spot in the video upload portal of the interactive page, or upload videos stored locally. The video content can be the user's self-introduction or a speech video made by the user on a certain speech topic. After the system obtains the uploaded video, it divides the audio of the video into frames of fixed length, calculates the MFCC features of each frame, and forms a feature vector sequence:

[0052] F = {f1, f2, ... f T};

[0053] Where T is the total number of frames; f T is the MFCC feature vector of the Tth frame.

[0054] In this embodiment, each frame of signal is subjected to a fast Fourier transform to obtain a spectrum, and the spectrum is filtered using a Mel filter bank; wherein the Mel filter bank is composed of a series of bandpass filters, the center frequencies of which are distributed according to the Mel scale, simulating the human ear's perception of different frequencies. A logarithmic operation is performed on the signal after passing through the Mel filter to obtain the logarithmic energy. The logarithmic energy is subjected to a discrete cosine transform, which converts the signal from the time domain to the frequency domain to obtain the MFCC coefficient, thereby representing the time-frequency characteristics of the speech signal, and the formula is:

[0055] MFCC = DCT(log(|STFT(x)| 2 ));

[0056] Among them, DCT is discrete cosine transform; STFT(x) is the short-time Fourier transform of the speech signal x.

[0057] Analyze MFCC features through AI technology to obtain scores corresponding to multiple specified eloquence dimensions; AI technology can use appropriate AI models, such as deep neural network (DNN), convolutional neural network (CNN), recurrent neural network (RNN) or long short-term memory network (LSTM), etc., use pre-collected training set data to train the specified AI model, and learn the relationship between MFCC features and eloquence dimension scores. Use the trained AI model to analyze the MFCC features of the speech videos uploaded by the user, and predict the scores corresponding to multiple specified eloquence dimensions.

[0058] Specifically, the fluency score can be obtained by analyzing the short-term energy and zero-crossing rate in the MFCC sequence and using the change of the short-term energy to detect pauses and non-fluency.

[0059] The pronunciation accuracy score can be calculated by comparing the MFCC features with a known accurate pronunciation template and calculating the distance between the MFCC features and the template. The smaller the distance, the higher the pronunciation accuracy score.

[0060] The speaking rate score can evaluate the speaking rate by measuring the number of phonemes or words per unit time, and use MFCC features to assist in the detection of phoneme boundaries.

[0061] Expressiveness can analyze the dynamic range, pitch changes, and rhythm changes in MFCC features, and use changes in pitch, volume, and duration to evaluate the expressiveness of speech and obtain an expressiveness score.

[0062] In addition, the vocabulary richness score can be based on the results of text analysis after speech-to-text conversion. NLP technology can be used to evaluate the diversity and complexity of vocabulary and obtain the corresponding vocabulary richness score.

[0063] The sentence complexity score can be based on text analysis to evaluate the complexity of sentence structure. It can analyze grammatical features such as sentence length, clause usage, passive voice, etc. to obtain a sentence complexity score.

[0064] The content relevance score can also be combined with NLP technology to perform semantic analysis on the voice content and evaluate the degree of match between the voice content and a given topic or question, thereby obtaining a content relevance score.

[0065] Similar to the process of analyzing the user's eloquence level based on the self-assessment answers and pushing corresponding eloquence resources, during the speech video analysis process, the video score is calculated based on the score of the designated eloquence dimension obtained from the speech video analysis and the weights corresponding to each designated eloquence dimension. The user's eloquence level is understood based on the video score, and eloquence resources with corresponding permissions are pushed to the user based on the user's eloquence level obtained from the video analysis alone, so that the user can conduct eloquence training and learning based on the eloquence resources that match his or her own eloquence level.

[0066] In this embodiment, in order to further improve the accuracy of eloquence resource recommendations, a multivariate regression analysis is performed based on the self-assessment question scores and the speech video scores to obtain a comprehensive score based on multiple specified eloquence dimensions. According to the comprehensive score, eloquence resources with corresponding permissions are recommended to improve the accuracy of eloquence resources.

[0067] like Figure 1 As shown, the eloquence resource recommendation method of this embodiment includes the following steps:

[0068] Step S1: Obtain the self-evaluation answers made by the user according to the preset questions, and preliminarily classify and score the self-evaluation answers according to the preset weights of each preset question to obtain the first score of each specified eloquence dimension.

[0069] Among them, the eloquence dimensions specified in this embodiment include fluency, pronunciation accuracy, speaking speed, vocabulary richness, sentence complexity, expressiveness, and content relevance. The calculation method of the first score is the same as the process of analyzing the self-assessment answers to obtain the preliminary eloquence score, and will not be repeated in detail here.

[0070] Step S2: Obtain the speech video uploaded by the user, perform audio feature analysis on the speech video to generate a feature vector sequence, and calculate the second score of each specified eloquence dimension based on the feature vector sequence.

[0071] The calculation method of the second score is the same as the above-mentioned method of obtaining the corresponding score of the specified eloquence dimension based on the MFCC features corresponding to the video and NLP technical analysis, and the detailed description will not be repeated here.

[0072] In addition, the method of obtaining eloquence dimension scores based on speech video analysis also includes:

[0073] According to the speech features of each frame in the speech video, the total pause duration and the average pause duration corresponding to each pause are counted;

[0074] The difference between the total pause duration and the average pause duration is calculated, and the rate of change of fluency over time is calculated based on the difference and a preset adjustment coefficient, and the fluency score is determined based on the rate of change.

[0075] Specifically, in this embodiment, the fluency change rate is evaluated by a calculus equation:

[0076]

[0077] in:

[0078] is the rate of change of fluency over time;

[0079] λ is the adjustment coefficient;

[0080] T pause is the total pause duration;

[0081] μ pause is the average pause duration.

[0082] That is, the total pause duration and the average pause duration corresponding to each pause are counted according to the speech features of each frame;

[0083] The difference between the total pause duration and the average pause duration is calculated, and the rate of change of fluency over time is calculated based on the difference and a preset adjustment coefficient, and the fluency score is determined based on the rate of change.

[0084] In addition, the method of obtaining eloquence dimension scores based on speech video analysis also includes:

[0085] Extracting first-dimension features and second-dimension features related to expressiveness based on speech videos;

[0086] Mapping the first dimension feature and the second dimension feature to the real part and the imaginary part of the complex number respectively;

[0087] The expressiveness score is determined by analyzing the modulus and phase of the complex number.

[0088] Specifically, the expressive force E is represented by a complex number z, which can be decomposed into a real part and an imaginary part: E=Re(z)+i*Im(z); wherein Re(z) is the real part of the complex number z, Im(z) is the imaginary part of the complex number z, and i is the imaginary unit.

[0089] Determine in advance two independent dimensions of expressiveness, such as volume and pitch, or rhythm and emotional intensity.

[0090] The speech signal corresponding to the speech video is analyzed, and the first dimension features and second dimension features related to expressiveness are extracted therefrom; the extracted first dimension features and second dimension features are converted into complex form. Each speech segment or speech sample can be represented by a complex number, that is, the first dimension features and the second dimension features are mapped to the real part and imaginary part of the complex number respectively. The real part can represent the volume or the pitch, and the imaginary part can represent the change of rhythm or the intensity of emotion.

[0091] The magnitude and phase of the complex number are analyzed to evaluate the expressiveness, thereby obtaining the corresponding expressiveness score. The magnitude of the complex number can be calculated to quantitatively evaluate the strength of the expressiveness, while the change in phase can reflect the changing trend or style of the expressiveness.

[0092] Step S3: Analyze the first score and the second score corresponding to each specified eloquence dimension based on the multiple regression model to obtain a comprehensive score.

[0093] The first score corresponding to each designated eloquence dimension is obtained by analyzing the self-evaluation answers obtained by the user answering the self-evaluation questions, and the second score corresponding to each designated eloquence dimension is obtained by analyzing the speech video uploaded by the user. In this embodiment, a multivariate regression model is used to perform a multivariate regression analysis on the first score and the second score corresponding to each designated eloquence dimension to obtain a comprehensive score.

[0094] In this embodiment, the first scores corresponding to the designated eloquence dimensions obtained by analyzing the answers to the self-assessment questions and the second scores corresponding to the designated eloquence dimensions obtained by analyzing the video are used as independent variables, and the comprehensive score is used as the dependent variable; a multiple regression model is established to determine the influence weight (regression coefficient) of each independent variable on the comprehensive score, and the regression coefficient obtained by the multiple regression model analysis is used to calculate the comprehensive score in combination with the first score and the second score. In this embodiment, the designated eloquence dimensions refer to fluency, pronunciation accuracy, speaking speed, vocabulary richness, sentence complexity, expressiveness and content relevance.

[0095] In a multiple regression model, evaluating the combined effect of a variable (such as pronunciation accuracy) on other dimensions (such as fluency, speaking rate, vocabulary richness, etc.) can be achieved by calculating the partial derivative of the variable. Higher-order partial derivatives can be used to measure interactions and nonlinear relationships between variables.

[0096] Specifically, calculate the first-order partial derivative of the dependent variable Y with respect to each independent variable Xi to obtain the marginal effect of each variable: This indicates the effect of the independent variable Xi on Y while keeping other variables constant.

[0097] Compute higher-order partial derivatives of the dependent variable Y with respect to some combinations of the independent variables to assess interactions between the variables:

[0098]

[0099] If this value is not zero, it indicates that there is an interaction between Xi and Xj.

[0100] If there is a nonlinear relationship or interaction between variables, an interaction term can be added to the multiple regression model. In this case, the multiple regression model expression is:

[0101] Y=β0+β1X1+β2X2+…+βijXiXj+∈

[0102] Among them, β0 is the intercept, β1, β2,…,βn are regression coefficients, ∈ is the error term, and βij is the coefficient of the interaction term. There is an interaction between Xi and Xj.

[0103] The dependent variable in this embodiment is the comprehensive score S composite , the independent variable combination is pronunciation accuracy score A accuracv and vocabulary richness score V vocabularv , then the interaction between variables is:

[0104]

[0105] Among them, γ is the influence coefficient.

[0106] During the model training process, the regression coefficients βi can be obtained through techniques such as the least squares method. These coefficients are the values ​​of partial derivatives, which reflect the contribution of each eloquence dimension to the comprehensive score.

[0107] Step S4: Select and push eloquence resources with corresponding permissions based on the comprehensive scores.

[0108] This embodiment can assign different roles and permissions to users based on their comprehensive scores, which can better manage and optimize system resources and ensure personalized and efficient user experience. The following are specific differences and examples of roles and permissions for high-scoring and low-scoring users:

[0109] When the comprehensive score is higher than the preset score, the user can be considered a high-score user, and the corresponding role is Advanced User; the corresponding permissions can be set as:

[0110] 1. Access to advanced training resources: Advanced users can access more advanced eloquence training videos and materials, including lectures by famous teachers, advanced presentation skills courses, etc.

[0111] 2. Participate in advanced discussion groups: Advanced users can join the advanced discussion groups in the system to conduct in-depth exchanges and discussions with other advanced users and experts.

[0112] 3. Get personalized coaching: Premium users can get personalized coaching and advice from the system, including weekly expert coaching.

[0113] 4. Publish and manage content: Advanced users can publish their own speech videos or eloquence training courses, and evaluate and manage other users' content.

[0114] 5. High-priority technical support: Advanced users can get higher-priority technical support and services when they encounter problems.

[0115] When the comprehensive score is lower than the preset score, the user can be considered a low-score user, and the corresponding role is Basic User; the permissions can be set as:

[0116] 1. Access basic training resources: Basic users can access basic public speaking training videos and materials, including entry-level presentation skills courses.

[0117] 2. Participate in basic discussion groups: Basic users can join basic discussion groups in the system to communicate and discuss with other basic users.

[0118] 3. Get system suggestions: Basic users can get basic suggestions and practice plans provided by the system, but there is no personalized tutoring service.

[0119] 4. Restricted content publishing: Basic users can upload their own speech videos for system evaluation, but cannot publish training courses or evaluate and manage other users' content.

[0120] 5. Standard technical support: Basic users can get standard technical support services when they encounter problems.

[0121] The following is an example:

[0122] High-scoring user: Zhang San

[0123] Zhang San scored high in the eloquence assessment, and his overall score \(S_{content}\) was 85 points, which makes him an advanced user.

[0124] -Access advanced training resources: Zhang San can access advanced eloquence training courses in the system, such as "Advanced Speech Skills and Practice".

[0125] - Participate in the advanced discussion group: Zhang San can join the advanced discussion group to have in-depth discussions with other advanced users and speech experts.

[0126] -Get personalized coaching: The system arranges expert coaching for Zhang San once a week to provide personalized speech skills advice.

[0127] -Publish and manage content: Zhang San can upload his own speech videos for other users to watch and learn from, and can also comment on other users' content.

[0128] -High-priority technical support: When Zhang San encounters a problem, he can get fast-response technical support through a dedicated channel for advanced users.

[0129] Low-scoring user: John

[0130] Li Si scored low in the eloquence assessment, and his overall score \(S_{content}\) was 45 points, which is a basic user.

[0131] - Access basic training resources: Li Si can only access basic eloquence training courses in the system, such as "Introduction to Primary Public Speaking Skills".

[0132] -Participate in the basic discussion group: Li Si can join the basic discussion group to exchange learning experiences with other basic users.

[0133] -Get system suggestions: The system provides Li Si with a basic practice plan, such as 30 minutes of reading aloud training every day.

[0134] -Restricted content publishing: Li Si can upload his own speech videos for system evaluation, but cannot publish training courses or manage other users' content.

[0135] -Standard technical support: When Li Si encounters a problem, he can obtain technical support through the standard user channel, but the response speed is slow.

[0136] This embodiment can achieve effective management of system resources and improve the personalization and pertinence of user experience by assigning different roles and permissions to users based on comprehensive scores. High-scoring users can obtain more resources and support to help them further improve their eloquence; low-scoring users can gradually improve through basic training, which motivates them to work hard and eventually achieve higher scores and more permissions.

[0137] The embodiment of the present invention provides an eloquence resource recommendation system, which executes the eloquence resource recommendation method of embodiment 1. The functions of each module in the system of the embodiment of the present invention can refer to the corresponding description in the above method, which will not be repeated here.

[0138] An embodiment of the present invention further provides an electronic device, Figure 2 FIG. 2 shows a structural block diagram of an electronic device according to an embodiment of the present invention. Figure 2 As shown, the electronic device includes: a memory 100 and a processor 200. The memory 100 stores a computer program that can be run on the processor 200. When the processor 200 executes the computer program, the method for recommending eloquence resources in the above embodiment is implemented. The number of the memory 100 and the processor 200 can be one or more.

[0139] The electronic device also includes:

[0140] The communication interface 300 is used to communicate with external devices and perform data exchange transmission.

[0141] If the memory 100, the processor 200 and the communication interface 300 are implemented independently, the memory 100, the processor 200 and the communication interface 300 can be connected to each other through a bus and communicate with each other. The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus. The bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 2 Only one thick line is used in the diagram, but this does not mean that there is only one bus or only one type of bus.

[0142] Optionally, in a specific implementation, if the memory 100, the processor 200 and the communication interface 300 are integrated on a chip, the memory 100, the processor 200 and the communication interface 300 can communicate with each other through an internal interface.

[0143] An embodiment of the present invention provides a computer-readable storage medium storing a computer program, which implements the method provided in the embodiment of the present invention when the program is executed by a processor.

[0144] An embodiment of the present invention further provides a chip, which includes a processor for calling and executing instructions stored in the memory from the memory, so that a communication device equipped with the chip executes the method provided by the embodiment of the present invention.

[0145] An embodiment of the present invention also provides a chip, including: an input interface, an output interface, a processor and a memory, wherein the input interface, the output interface, the processor and the memory are connected via an internal connection path, and the processor is used to execute the code in the memory. When the code is executed, the processor is used to execute the method provided by the embodiment of the invention.

[0146] It should be understood that the processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor, etc. It is worth noting that the processor may be a processor supporting the advanced RISC machines (ARM) architecture.

[0147] Further, optionally, the above-mentioned memory may include a read-only memory and a random access memory, and may also include a non-volatile random access memory. The memory may be a volatile memory or a non-volatile memory, or may include both volatile and non-volatile memories. Among them, the non-volatile memory may include a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), or a flash memory. The volatile memory may include a random access memory (RAM), which is used as an external cache. By way of example but not limitation, many forms of RAM are available. For example, static RAM (SRAM), dynamic random access memory (DRAM), synchronous DRAM (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link dynamic random access memory (SLDRAM) and direct rambus RAM (DR RAM).

[0148] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function according to the present invention is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium.

[0149] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine different embodiments or examples described in this specification and the features of different embodiments or examples, unless they are contradictory.

[0150] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In the description of the present invention, the meaning of "plurality" is two or more, unless otherwise clearly and specifically defined.

[0151] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of various changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A method for recommending eloquence resources, characterized in that: include: Obtaining the self-evaluation answers made by the user according to the preset questions, and preliminarily classifying and scoring the self-evaluation answers according to the preset weights of each of the preset questions to obtain a first score for each designated eloquence dimension; Obtaining a speech video uploaded by a user, performing audio feature analysis on the speech video to generate a feature vector sequence, and calculating a second score for each of the specified eloquence dimensions based on the feature vector sequence; Analyze the first score and the second score corresponding to each of the specified eloquence dimensions based on a multiple regression model to obtain a comprehensive score; According to the comprehensive score, eloquence resources of corresponding authority are selected and pushed.

2. The method for recommending eloquence resources according to claim 1, characterized in that: The specified dimensions of eloquence include fluency, pronunciation accuracy, speaking rate, vocabulary richness, sentence complexity, expressiveness, and content relevance.

3. The method for recommending eloquence resources according to claim 1, characterized in that: The step of performing audio feature analysis on the speech video to generate a feature vector sequence includes: The audio data corresponding to the speech video is divided into frames of fixed length, and the speech features of each frame are extracted using Mel-frequency cepstral coefficients to form a feature vector sequence.

4. The method for recommending eloquence resources according to claim 1, characterized in that: The step of calculating a second score for each designated eloquence dimension based on the feature vector sequence includes: The feature vector sequence is analyzed based on a pre-built deep neural network to obtain the second score of the specified eloquence dimension.

5. The method for recommending eloquence resources according to claim 2, characterized in that: Also includes: Counting the total pause duration and the average pause duration corresponding to each pause according to the speech features of each frame in the speech video; The difference between the total pause duration and the average pause duration is calculated, the change rate of the fluency over time is calculated according to the difference and a preset adjustment coefficient, and the fluency score is determined according to the change rate.

6. The method for recommending eloquence resources according to claim 2, characterized in that: Also includes: Extracting first dimension features and second dimension features related to expressiveness based on the speech video; Mapping the first dimensional feature and the second dimensional feature to the real part and the imaginary part of a complex number respectively; The expressiveness score is determined by analyzing the magnitude and phase of the complex number.

7. The method for recommending eloquence resources according to claim 1, characterized in that: The calculation method of the comprehensive score includes: The first scores corresponding to the designated eloquence dimensions and the second scores corresponding to the designated eloquence dimensions are used as independent variables of the multiple regression model, and the comprehensive scores are used as dependent variables of the multiple regression model; Determine the regression coefficient of each of the independent variables on the comprehensive score, and perform a multivariate regression analysis on the first scores corresponding to each of the specified eloquence dimensions and the second scores corresponding to each of the specified eloquence dimensions using the regression coefficients to obtain the comprehensive score.

8. An eloquence resource recommendation system, characterized in that: Execute the eloquence resource recommendation method as described in any one of claims 1 to 7.

9. An electronic device, characterized in that: include: A processor and a memory, wherein the memory stores instructions, and the instructions are loaded and executed by the processor to implement the eloquence resource recommendation method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method for recommending eloquence resources according to any one of claims 1 to 7 is implemented.

Images