CN114254182A

CN114254182A - Data processing method and device for learning platform and computer system

Info

Publication number: CN114254182A
Application number: CN202011001217.6A
Authority: CN
Inventors: 杨志豪
Original assignee: Wencheng Intelligent Education Co ltd
Current assignee: Wencheng Intelligent Education Co ltd
Priority date: 2020-09-22
Filing date: 2020-09-22
Publication date: 2022-03-29

Abstract

The application discloses a data processing method, a device and a computer system of a learning platform, which comprises a data processing method of an online learning platform, wherein the method comprises the following steps: acquiring user data and a target learning category corresponding to a user, wherein the user data comprises course learning records of the user; predicting the course characteristics preferred by the user by adopting a preset algorithm according to the user data; acquiring course characteristics corresponding to the courses of the target learning category; determining the similarity between the user and the courses according to the course characteristics preferred by the user and the course characteristics corresponding to the courses of the target learning category; and determining the course as the target course of the user based on the similarity and the preset condition so as to return the target course to the user, so that the user can conveniently find the video matched with the requirement, and the efficiency and the accuracy of course matching are improved.

Description

Data processing method and device for learning platform and computer system

Technical Field

The invention relates to the technical field of computers, in particular to a data processing method and device of a learning platform and a computer system.

Background

With the development of internet technology, online education, which is a supplement to modern conventional education, is increasingly developed in large scale. However, because there are a large number of courses on the online education platform, the user often needs to spend a lot of time on browsing the contents of each course to find a suitable course, but still cannot guarantee that the course matching the user's needs can be found, which not only consumes time, but also cannot guarantee that the course found by the user can meet the user's needs.

Disclosure of Invention

In order to solve the defects of the prior art, the main object of the present invention is to provide a data processing method, device and computer system for a learning platform.

In order to achieve the above object, in one aspect of the present invention, there is provided a data processing method of a learning platform, the method including:

acquiring user data and a target learning category corresponding to a user, wherein the user data comprises course learning records of the user;

predicting the course characteristics preferred by the user by adopting a preset algorithm based on the user data;

acquiring course characteristics corresponding to the courses of the target learning category;

determining similarity between the user and the courses based on the course characteristics of the preference of the user and the course characteristics corresponding to the courses of the target learning category;

and determining the course to be the target course of the user to return to the user based on the similarity and a preset condition.

In some embodiments, the course learning record of the user includes the historical learning time of the user and the historical course score corresponding to the historical learning time, and the method further includes:

and predicting the target learning time of the user according to the historical learning time and the corresponding historical course score.

In some embodiments, the method further comprises: determining a demonstration skill evaluation of the teacher based on teacher videos corresponding to the target learning category and determining the lesson to be a target lesson for the user based on the demonstration skill evaluation of the teacher, wherein

Said determining a demonstration skill rating further comprises:

extracting image frames of the course video;

generating a demonstration skill evaluation of the teacher based on the input preprocessed image frames first preset model.

In an embodiment of the invention, the teacher video is a short video, with a maximum of 3 minutes. The short video may be given the most accurate or relevant knowledge.

Through the short video, the platform system can evaluate the demonstration skill of the teacher and can create an index corresponding to the teacher, so that the course desired by the user can be selected more effectively.

In addition, the platform system may also create "notes" corresponding to the short videos to store personalized notes or knowledge of the user for better learning or teaching.

In some embodiments, the first pre-set model includes a convolutional layer, a pooling layer, and a fully-connected layer.

In some embodiments, determining the teacher's tone rating based on a teacher video corresponding to the target learning category and determining the lesson as the user's target lesson based on the teacher's tone rating, wherein the determining the tone rating further comprises:

extracting the audio of the course video;

generating a tone rating for the teacher based on the input audio second preset model.

In some embodiments, each of the lessons comprises a corresponding lesson video, the lesson features comprise keywords corresponding to the lesson, and the method further comprises: extracting the keywords, the extracting keywords further comprising:

acquiring a transcribed text of the course video by performing character transcription on the course video;

and filtering the transcribed text by using a preset NLP algorithm to generate a keyword corresponding to the course.

In some embodiments, the method further comprises:

acquiring a learning state image of the user when learning the target course;

predicting the emotional orientation of the user to the target course by adopting a third preset model based on the learning state image;

generating a course study record including the emotional orientation and the target course.

By employing the emotional orientation learned by the user, the user's attention and concentration may be focused, thereby better selecting courses of interest to the user.

In some embodiments, the determining the similarity between the user and the lesson according to the lesson features of the user's preference and the lesson features corresponding to the lesson of the target learning category further includes:

and determining the cosine similarity between the user and the course according to the course characteristics preferred by the user and the course characteristics corresponding to the course.

In some embodiments, the user data comprises a search record of the user, the method comprising:

receiving a search request of a user, wherein the search request comprises search terms;

predicting search keywords of the user according to the search words by using a fourth preset model;

and generating a search record of the user according to the search keyword.

In some embodiments, the lesson features include a corresponding specialty category for the instructor, the method further comprising:

acquiring a problem solution record of a teacher;

and predicting the professional category of the teacher according to the problem solution records by using a fifth preset model.

In some embodiments, the lesson features include corresponding preset geographic areas, the method further comprising:

acquiring the real-time position of the user;

determining the course to be a target course of the user to return to the user based on the relation between the real-time position and the preset geographic area.

In a second aspect of the present invention, there is provided a data processing apparatus of a learning platform, the apparatus comprising:

the system comprises an acquisition module, a processing module and a display module, wherein the acquisition module is used for acquiring user data and target learning categories corresponding to a user, and the user data comprises course learning records of the user; acquiring course characteristics corresponding to the courses of the target learning category;

the prediction module is used for predicting the course characteristics preferred by the user by adopting a preset algorithm according to the user data;

the matching module is used for determining the similarity between the user and the courses according to the course characteristics of the preference of the user and the course characteristics corresponding to the courses of the target learning category;

and the processing module is used for determining the course as the target course of the user to return to the user based on the similarity and a preset condition.

In a third aspect of the present invention, there is also provided a computer system comprising:

one or more processors; and memory associated with the one or more processors for storing program instructions that, when read and executed by the one or more processors, perform operations comprising:

predicting the course characteristics preferred by the user by adopting a preset algorithm according to the user data;

In a fourth aspect of the present invention, there is also provided a computer storage medium having computer-readable program code stored thereon, the code being executed by a processor to perform the data processing method of the above learning platform.

According to the embodiment of the invention, the characteristics of the courses interested by the user can be predicted according to the course learning record of the user, and the courses with the similarity meeting the conditions with the characteristics interested by the user are recommended to the user, so that the user can conveniently find the videos matched with the requirements, and the efficiency and the accuracy of course matching are improved. And the desired lessons can be acquired more precisely based on the emotional orientation of the user.

Drawings

In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings needed to be used in the description of the embodiments will be briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without creative efforts.

FIG. 1 is a flow chart of a data processing method for an educational platform provided by an embodiment of the present application;

FIG. 2 is a block diagram of a data processing device of an educational platform according to an embodiment of the present application;

fig. 3 is a computer system structure diagram provided in the embodiment of the present application.

Detailed Description

In order to make the objects, technical solutions and advantages of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.

As described in the background, in order to recommend a course to a user that meets the user's needs, the present application proposes an online education system that can provide targeted course recommendations. The system can be arranged on an online education platform, and users of the system comprise users for learning courses and teachers giving courses. The teacher can open and establish the course to upload the corresponding course video to on-line learning platform, the user can select corresponding course to study according to the demand. Each course has characteristics of corresponding price, teacher, course score, playing amount, and the like.

The system includes a database for storing user data and course learning records of each user. For example, the user profile includes a user ID, gender, age, target learning category, target learning level, and domain in which the user is located. The domain represents a user occupation or a domain in which the user is located. For example, if the user is a company software developer, the corresponding domain is IT.

In embodiments of the present invention, the database may also store a course study record for the user. The course learning record includes a user ID, a course name, a learning level of the course, a learning time, a course ID, and the like. In the database, characteristics of each course and corresponding course video are also stored. The characteristics of the lesson may include the language of the lesson, the amount of viewing of the lesson, the channel to which the lesson belongs, the tags for the lesson, the learning level for the lesson, the class to which the lesson belongs, the age of the instructor, the gender of the instructor, the assessment of the instructor's presentation skills, the tone assessment of the instructor, the specialty category of the instructor, the assessment of the lesson, the price of the lesson, and so forth.

The system includes a plurality of AI modules that may each be used to implement different functions.

Specifically, the AI module may include:

1. a search identification module: the module is used for identifying the content to be searched input by the user according to the previous search or input content of the user.

The search identification module may include an LSTM neural network model and predict the user's content to be searched using the trained LSTM neural network model based on the user's historical search history and interests, etc. Long Short Term Memory (LSTM) is an artificial Recurrent Neural Network (RNN) used in the field of deep learning. Unlike standard feedforward neural networks, LSTM has a feedback connection. It can handle not only a single data point (e.g., an image), but also an entire data sequence (e.g., voice or video). For example, LSTM is suitable for tasks such as unsegmentation in network communications or IDS (intrusion detection system), handwriting recognition of connections, speech recognition and anomaly detection.

Preferably, the module may predict keywords to be searched by the user based on the content that the user has entered. For example, the user enters a word from which the module can predict the complete keyword, such as the course name, that the user would like to search.

In an embodiment of the invention, the prediction process may comprise:

generating a corresponding tagged vector according to a search word input by a user;

creating a corpus containing all courses in a database;

creating a filling sequence for the corpus vectors generated according to the corpus;

and inputting the marked vector machine and the corpus vector into a preset model, and generating a predicted keyword to be searched by the user according to a preset prediction variable number.

Preferably, the corpus includes names of courses and descriptions of courses. Preferably, the number of predicted keywords is 5. Preferably, the description of the lesson may include characteristics of the lesson. Preferably, the model may comprise an LSTM model comprising an N-Gram algorithm, and the first layer of the LSTM model may be provided as an embedding layer.

Preferably, the module may include LSTM neural network models, Keras, N-Gram Algorithm, etc. machine learning models, and use these trained models to recognize the content of the user's handwriting/speech input, predicting what the user wants to search for.

For example, when a user inputs a keyword to be searched through voice, the module may recognize the voice to obtain a corresponding keyword.

For example, when the user inputs a keyword to be searched by handwriting, the module may identify and obtain a corresponding keyword according to a handwritten image.

2. A keyword extraction module: the module is used for extracting texts in the video, and keywords of corresponding courses can be determined through analysis according to the extracted texts. The labels of the videos and the corresponding learning categories can be determined according to the keywords.

Preferably, the Google Transcription API can be used to transcribe the video, extract the text in the video on the fly, and then filter the extracted text by using a preset NLP algorithm to obtain the course keyword. The obtained course keywords can be saved in a database so as to be used when needed.

After the teacher uploads the course video to the online education platform, the keyword extraction module can extract keywords of the course video, and according to the extracted keywords, the video theme of the course video uploaded by the teacher can be identified. According to the determined course theme, the courses can be classified into corresponding course categories and/or learning categories, and the video index is generated according to the video theme.

In an embodiment of the present invention, the selection of a more appropriate lesson for the instructor is facilitated by using the index of the video. Preferably, through the index of the video, the demonstration skill of the teacher can be effectively improved.

In an embodiment of the invention, the teacher may also answer the user's question in a forum on the online education platform. So that the professional category to which the teacher belongs can be determined based on the teacher's solution using the module. Preferably, the module can predict the teacher's professional class from the solution using a trained machine learning model such as a Convolutional Neural Network (CNN). Preferably, the characteristics of the lesson may include a specialty category determined by the module.

In an embodiment of the invention, for videos uploaded by teachers, the module can use a convolutional neural network model to identify emotions and speaking skills of teachers recording the video of courses. The module can generate the evaluation of the lecture skill of the teacher according to the image frames intercepted from the video and the extracted voice. The module may send the generated demonstration skill evaluations to the corresponding teacher for the teacher to perform the improvement of the demonstration skills based on the demonstration skill evaluations. A trained convolutional neural network model (CNN) may predict teacher's emotion and/or speech skills from the image frames.

Preferably, images can be extracted from the video using OpenCV to predict teacher's emotion and/or speech skills from the image frames.

The trained CNN model includes convolutional layers, pooling layers, and fully-connected layers.

The convolutional layer may perform a convolution operation on the image to extract corresponding convolution features. The convolutional layer is used to extract high-order features from the input image. The CNN model may include a plurality of convolutional layers, where the first convolutional layer is used to extract low-order features, such as edges, colors, and gradient directions of the image. Other convolutional layers enable the model to be used to extract higher order features, thereby providing a model that can fully extract image features. For single channel images, the convolution layer includes a filter that resolves the image into corresponding convolution features based on a preset amplitude distribution value. For a multi-channel image, the depth of the filter may be preset to be the same as the image.

Two results were obtained after convolutional layer operation: one is a reduction in the dimensionality of the convolution features and the other is an increase or a maintenance. The image may be effectively padded (Valid Padding) to decrease the latitude or the Same Padding (Same Padding) to the image increases or remains the Same for the latitude of the convolution feature.

The pooling layer is used to reduce the spatial size of convolution features generated by the convolutional layer. The pooling layer reduces the computational power required by the model to process the convolution features by performing dimensionality reduction on the convolution features. Meanwhile, the pooling layer can also be used for extracting the main characteristics of rotation and position invariance contained in the image so as to improve the efficiency of model training.

The pooling layer may include maximum pooling or average pooling. Maximum pooling may be used to obtain a maximum in the portion of the image covered by the pooling layer, and noise suppression operations may be taken while reducing dimensions. Average pooling may take the average of all values of the image portion covered by the pooling layer, taking dimensionality reduction as a noise suppression operation.

The fully-connected layer is used to learn combinations of higher-order features represented by the output of the convolutional layer, and the fully-connected layer may learn non-linear functions that may be present in the higher-order features represented by the output of the convolutional layer. The input image after passing through the convolutional layer, the pooling layer and the fully connected layer is input to a multi-order Perceptron (Perceptron) and then flattened (flatten) into column vectors. The flattened image may be input to a feed-forward neural network and backpropagated for each training iteration. After multiple times of training, the model can distinguish the main features and the low-level features included in the image, and then the model can determine the corresponding classification of the input image by using a Softmax technology.

The demonstration skill evaluation comprises evaluation of the body language, expression and intonation of the teacher. Preferably, the process of acquiring the teacher's demonstration skill evaluation by the module comprises:

s1, converting the intercepted image frame with RGB into a gray image;

preferably, one or more RGB-bearing image frames may be taken from the video.

S2, converting the gray level image into an array vector corresponding to the image frame;

s3, transmitting the array vector to a trained model, and predicting the corresponding score of each image frame according to the array vector by the trained model;

preferably, the trained model comprises a trained convolutional neural network model.

Preferably, the trained convolutional neural network model identifies teacher's motion within the image, and assigns a value of 0 to the image if adverse behavior (e.g., tongue protrusion, nose grasping, etc.) occurs. Frames that do not contain an action will be assigned a value of 2. The number of 1 s and 0 s will be calculated and the ratio of 1 s and 0 s will be used to predict how good the instructor's demonstration skill will be.

When the score of a certain image is not less than 6, the image is a bad image/bad frame. When bad frames occur continuously, the score or count of the bad frames will increase at a preset rate rather than the actual count.

The trained model may have an error constraint of +/-5% to +/-7%, i.e., a video that evaluates superior over 75% of the demonstration skills may be no or little misbehavior.

At the same time, the trained model presents an error that mistakes normal gestures as bad behavior, but the probability of this error is only 5%.

The teacher demonstration skill suggestion module may calculate the ratio of 1 to 0 for the image frames of the video based on the corresponding assignments for each image output by the trained model. When the ratio of 0 exceeds a preset threshold, the teacher's demonstration skill is evaluated as poor, and when the ratio does not exceed the preset threshold, the teacher's demonstration skill is evaluated as excellent.

Preferably, the module predicts the teacher's pitch rating based on the teacher's uploaded video. The tonal assessment may include a degree of softness and a degree of harshness.

The module can determine tone evaluation of the teacher according to the pitch, definition, strength and the like of the voice of the teacher in the video uploaded by the teacher. Due to two basic properties of sound: sampling frequency and audio signal value, pitch evaluation can be obtained by the following steps:

according to the audio extracted from the course video, calculating a sampling frequency and an audio signal corresponding to the audio by using a preset method;

calculating the average value of the audio signal of the audio according to the audio signal;

acquiring the ratio of the average value of the audio signal to the average sampling frequency of the audio;

when the ratio is greater than 1, it indicates that the audio is soft, low in decibels, and not sharp, and thus the degree of softness of the audio is 100 and the degree of harshness is 0. And when the pitch is too high, the ratio will be negative, which means that the audio is harsh. For example, for a ratio in the range of-0.3 to-0.5, the audio may be considered to be 50 soft and 50 harsh.

Determining a pitch rating for the audio based on the ratio.

Preferably, the module may preset a pitch rating corresponding to the value range of each ratio to determine the pitch rating of each audio. The module may determine a corresponding teacher's pitch rating based on the pitch rating of the audio.

3. A tag search module: the module may be operable to search for a corresponding course based on the user-selected tag.

The tags may be predetermined using NLP natural language processing techniques. For example, the corresponding tag may be extracted in advance from the title of the lesson video, the text content obtained by transcription, and the like.

The tag determination process includes a preprocessing process of text data using NLP natural language processing technology, i.e., converting the data into contents that can be understood by a computer. One of the main processes of preprocessing is to filter out useless data, including common words that have been ignored by programming (e.g., "the", "a", "an", "in", etc.).

From the preprocessed text data, tags describing the content or subject of the lesson, etc. may be generated.

4. A recommendation engine module: the module may recommend courses to the user based on the user's physical and emotional behavior, i.e., emotional orientation, or based on the user's user profile and course learning records.

Preferably, the course recommendation may be performed using the following steps:

s1, acquiring the target learning type, user data and course learning record of the user;

s2, predicting the course characteristics preferred by the user according to the target learning type, the user data and the course learning record;

specifically, the user data further includes the use conditions of the online learning platform of the user, including data of browsed courses and browsing times, browsed forum problems and browsing times, types of the main browsed courses, types of teachers with the largest number of contact times, and the like.

The recommendation module vectorizes the user profile using tfidf based on the Ngrams model to determine the characteristics of the user.

The Ngrams model is a natural language processing model that can be used to extract keywords from user profiles.

TFIDF is a statistical method that can be used to count how important a word is to a document.

And the recommendation engine module acquires the course characteristics of the courses according to the courses contained in the course learning record of the user, and predicts the preferred course characteristics of the user by using the trained machine learning model according to the characteristics and the course characteristics of the user.

Specifically, the machine learning model has been trained in advance using a training data set, which includes user data and course learning records of other users.

S3, obtaining course characteristics corresponding to the courses of the target learning category;

specifically, the user may select a target learning category for which learning is desired by searching or clicking on the learning category option. Specifically, the learning categories include a learning major category and a learning minor category, each learning major category includes one or more learning minor categories, and the user may select the learning major category or the learning minor category as the target learning category.

S4, determining the similarity between the user and the courses of the target learning category according to the course characteristics preferred by the user, the course characteristics corresponding to the courses of the target learning category and the preset weight of each course characteristic.

Specifically, the recommendation engine module determines in advance a weight of each course feature, which can be determined according to the course learning records of all users.

The recommendation engine module can determine similarity between the course characteristics preferred by the user and the course characteristics corresponding to the courses of the target learning category by adopting algorithms such as word vector similarity. Preferably, a cosine similarity algorithm may be used for calculation.

The word vector similarity algorithm may calculate the similarity between the generated vectors based on the vectors of a given text block. The word vector similarity algorithm comprises algorithms such as cosine similarity, Euclidean distance, Jacard distance, word shifter distance and the like.

Cosine similarity algorithm is an algorithm that can be used to measure the similarity of documents or data.

And S5, recommending the target courses with the similarity meeting the preset conditions to the user.

The recommendation module may recommend one or more target courses to the user for selection by the user.

In particular, the recommendation module may also use other recommendation algorithms or methods.

Preferably, the recommending module may further obtain a real-time location of the user, each course includes a corresponding preset geographic tag, and the recommending module may recommend the course to the user, where the corresponding geographic tag matches the real-time location of the user. The fact that the geographic label is consistent with the real-time position of the user comprises that the geographic area represented by the geographic label contains the real-time position of the user and/or the distance between the geographic area represented by the geographic label and the real-time position of the user is smaller than a preset value. Preferably, the recommendation module can obtain the real-time location of the user through the Google Places API to recommend the course to the user.

The user may also upload images to the course platform system, and the recommendation module may predict objects contained in the images and recommend courses to the user that tag the objects contained therein.

For example, a user can upload an image of a mobile phone to the platform, the recommendation module predicts that the probability that an object corresponding to the image is the mobile phone is 80% and the probability that the object is the automobile is 60% by using a preset AI image recognition module, then obtains courses with labels including the mobile phone, the gadget or the automobile, and displays the courses to the user according to the probability sequence.

Preferably, the recommending module may further preset predefined recommending rules, and the recommending module may recommend the courses to the user according to the recommending rules.

For example, the recommendation rules include: a user who has learned Python will learn Python for Web development. The recommendation module will recommend Python for Web development to users who have learned Python. Preferably, the recommendation module may generate the recommendation rule by using a preset algorithm according to the user data and the course learning record of the user.

5. A learning process module: the module may be used to record a learning history of a user, who uses the module to select a level of learning desired. The user's lesson learning records may also be stored by the module. Each course has a corresponding learning level and learning category.

Preferably, the process of creating the learning history of the user includes:

receiving a target learning level selected by a user;

receiving a target learning category selected by a user;

and matching courses which accord with the target learning grade and the target learning category, and sequencing according to the playing times or the number of people who like the courses.

The user can select the course in the courses which accord with the target learning grade and the target learning category for learning, and the module records the course learning record of the user.

6. Course comparison module: the module is used for comparing a plurality of courses for the user when the plurality of courses are recommended to the user by the modules such as the recommendation engine or the learning process module. Such a comparison may be based on parameters such as price of the lesson, number of views, number of times enjoyed by the user, lesson score, etc., and recommend the most appropriate lesson for the user based on these parameters.

7. And an emotion tracking module: the module can be used for identifying the shot images of the user when watching the lesson video and determining the emotional orientation of the user to the video being watched. Preferably, the emotion curve may include like, dislike, etc. Image frames in the video may be extracted and the facial and body movements, etc. of the user contained in the images may be identified using a convolutional neural network model (CNN), etc. to predict the user's emotional orientation to the lesson.

Preferably, images can be extracted from the video using OpenCV to perform recognition of the facial and body movements and the like of the user included in the images. OpenCV is an open source technology that can be used for image processing.

8. An optimal learning time module:

the module is used for predicting the target learning time of the user according to the historical course learning record of the user. The module predicts the target learning time that the user may obtain the best learning effect according to the historical learning time included in the historical course learning record and the historical course score corresponding to each historical learning time.

For example, the module may predict a target learning time for a user based on respective lesson scores for the user when learning the same lesson over three different time periods.

Preferably, the convolutional neural network model is used, the historical learning time of the user and the corresponding historical lesson score are used as a training data set, the learning time of the user which is possible to obtain the highest lesson score is predicted, and the learning time is recommended to the user as the target learning time, so that the user can learn at a proper time, and the learning efficiency is improved.

The online learning system can be used for recommending courses for users, and the process comprises the following steps:

step one, acquiring a target learning category of a user;

the online learning system may determine a target learning category of the user based on the learning category selected by the user. The user can input the keywords of the learning category to be learned through keyboard input or handwriting and the like, and the online learning system predicts the target learning category of the user by using the search recognition module. The user may also send the target learning category directly to the online learning system. The online learning system can also acquire a target learning category preset by the user from the learning process module.

Step two, acquiring user data and course learning records of a user;

the online learning system acquires the stored user data and the course learning record from the learning process module. The user data includes user ID, usage of the online learning platform of the user, browsed courses and browsing times, browsed forum problems and browsing times, types of the main browsed courses, types of teachers with the largest number of contact times, and the like.

The course learning record includes user ID, course name, learning level of the course, learning time, and course ID. Preferably, the lesson learning record further comprises the emotional orientation of the user to the lesson that has been learned.

Step three, the online learning system calls a recommendation module to predict the course characteristics preferred by the user;

preferably, the recommendation module can vectorize the user profile using tfidf based on the Ngrams model to extract the user's preferred course characteristics from the user profile and the course learning records.

Step four, acquiring course characteristics corresponding to the courses of the target course category;

each course has corresponding course characteristics, and the course characteristics can comprise data such as the gender and the like of the teacher recorded by the teacher, and also can comprise demonstration skill evaluation and tone evaluation of the teacher identified by the teacher demonstration skill suggestion module, keywords of the course extracted by the keyword extraction module, and theme labels of the course determined by the teacher video theme determination module.

Step five, determining the similarity between the user and each course according to the course characteristics preferred by the user and the course characteristics corresponding to the courses of the target learning category;

preferably, the similarity between the user and the course can be determined according to the calculated course characteristics of the preference of the user and the cosine similarity of the course characteristics corresponding to the course of the target learning category.

Recommending target courses with similarity meeting preset conditions to a user;

preferably, the preset number of courses with the highest similarity can be determined as the target course, and the target course is pushed to the user.

The number of the target courses can be not less than 2, and the course comparison module can predict the most suitable courses for the user based on the parameters of the prices, the watching times, the times enjoyed by the user, the course scores and the like of the courses.

The online learning system generates a recommendation list containing all the targeted courses, including the most appropriate marked courses, to send to the user.

Preferably, the recommendation list may include the course video of the target course and/or an access link to the course video of the target course.

Preferably, the online learning system may predict a target learning time corresponding to the user by using the optimal learning time module, where the target learning time is a predicted learning time for which the probability that the lesson score is the highest when the user learns at the target learning time is obtained.

The online learning system can use an emotion tracking module to shoot images of the user when watching the course video and determine the emotion orientation of the user on the course video. The learning history module can store the emotional orientation and the course scores of the user during learning.

The online learning system may also generate demonstration skill evaluations for the teacher, the process comprising:

A. the teacher uploads the course video to an online learning system;

upon uploading the lesson video, the teacher may enter the selected learning category and other lesson characteristics for the lesson into the online learning system.

The keyword extraction module can determine the subject of the course according to the uploaded course video. When the learning category selected by the teacher does not match the course theme, the learning category of the course can be modified into the learning category corresponding to the course theme.

B. Carrying out image frame extraction on the uploaded curriculum video, and generating demonstration skill evaluation and tone evaluation on a teacher by using a teacher demonstration skill suggestion module according to the extracted image frames;

C. sending the generated demonstration skill evaluation and tone evaluation to a teacher;

the teacher can adjust according to the demonstration skill evaluation and the tone evaluation received to promote the quality of giving lessons.

Corresponding to the above embodiments, the present invention provides a data processing method for an online platform, as shown in fig. 1, the method includes:

110. acquiring user data and a target learning category corresponding to a user, wherein the user data comprises course learning records of the user;

120. predicting the course characteristics preferred by the user by adopting a preset algorithm according to the user data;

130. acquiring course characteristics corresponding to the courses of the target learning category;

140. determining the similarity between the user and the courses according to the course characteristics preferred by the user and the course characteristics corresponding to the courses of the target learning category;

preferably, the determining the similarity between the user and the course according to the course characteristics of the preference of the user and the course characteristics corresponding to the course of the target learning category includes:

141. and determining the cosine similarity between the user and the course according to the course characteristics preferred by the user and the course characteristics corresponding to the course.

150. And determining the corresponding course with the similarity meeting the preset condition as the target course of the user and returning the target course to the user.

Preferably, the course learning record of the user includes historical learning time of the user and a historical course score corresponding to each historical learning time, and the method includes:

160. and predicting the target learning time of the user according to the historical learning time and the corresponding historical course score.

Preferably, each course includes a corresponding course video and a teacher, the course characteristics include demonstration skill evaluation of the teacher, and the determination process of the demonstration skill evaluation includes:

161. extracting image frames of the course video;

162. generating a demonstration skill evaluation of the teacher based on the input pre-processed image frame first preset model.

Preferably, the lesson features include a tone evaluation of the teacher, and the determining of the tone evaluation includes:

163. extracting the audio of the course video;

164. based on the input audio, a second pre-set model generates a tone rating for the teacher.

Preferably, the course characteristics include keywords corresponding to the course, and the keyword extraction process includes:

165. performing character transcription on the course video to obtain a transcription text of the course video;

166. and filtering the transcribed text by using a preset NLP algorithm to generate a keyword corresponding to the course.

Preferably, the method comprises:

167. visually presenting the target course to the user, and collecting a learning state image when the user learns the target course;

168. predicting the emotion orientation of the user on the target course by adopting a third preset model according to the learning state image;

169. and generating and storing a course learning record containing the emotion orientation and the target course.

Preferably, the user data includes a search record of the user, and the method includes:

170. receiving a search request of a user, wherein the search request comprises search terms;

171. predicting search keywords of the user according to the search words by adopting a fourth preset model;

172. and generating and storing the search record of the user according to the search keyword.

Preferably, the lesson features include a corresponding specialty category of the instructor, and the method includes:

173. acquiring a problem solution record of a teacher;

174. and predicting the professional category of the teacher according to the problem solution records by using a fifth preset model.

Preferably, the lessons include corresponding preset geographic areas, and the method includes:

175. acquiring the real-time position of the user;

176. determining the course to be a target course of the user to return to the user based on the relation between the real-time position and the preset geographic area.

In an embodiment of the present invention, the present application further provides a data processing apparatus of an online platform, as shown in fig. 2, the apparatus includes:

an obtaining module 210, configured to obtain user data and a target learning category corresponding to a user, where the user data includes a course learning record of the user; acquiring course characteristics corresponding to the courses of the target learning category;

the prediction module 220 is configured to predict, according to the user data, a course characteristic preferred by the user by using a preset algorithm;

a matching module 230, configured to determine similarity between the user and the course according to the course characteristics of the preference of the user and the course characteristics corresponding to the course of the target learning category;

and the processing module 240 is configured to determine that the corresponding course with the similarity meeting the preset condition is the target course of the user and returns the target course to the user.

Preferably, the matching module 230 is further configured to predict a target learning time of the user according to the historical learning time and the corresponding historical lesson score.

Preferably, the apparatus further comprises a generating module 250, configured to perform image frame extraction on the curriculum video; generating a demonstration skill evaluation of the teacher based on the input preprocessed image frames first preset model.

Preferably, the lesson features include a tone rating of the instructor, and the generation module 250 is further operable to extract audio of the lesson video; generating a tone rating for the teacher based on the input audio second preset model.

Preferably, the course characteristics include keywords corresponding to the course, and the generating module 250 is further configured to perform text transcription on the course video to obtain a transcribed text of the course video; and filtering the transcribed text by using a preset NLP algorithm to generate a keyword corresponding to the course.

Preferably, the processing module 240 is further configured to visually present the target course to the user, and collect a learning state image of the user when learning the target course; the prediction module 250 is further configured to predict, according to the learning state image, an emotional orientation of the user to the target course by using a third preset model; and generating and storing a course learning record containing the emotion orientation and the target course.

Preferably, the matching module 220 is further configured to determine the cosine similarity between the user and the course according to the course characteristics of the user's preference and the course characteristics corresponding to the course.

Preferably, the user data includes a search record of the user, and the generating module 250 is further configured to receive a search request of the user, where the search request includes a search term; predicting search keywords of the user according to the search words by adopting a fourth preset model; and generating and storing the search record of the user according to the search keyword.

Preferably, the lesson features include corresponding professional categories of the teacher, and the generating module 250 is further configured to obtain a question answering record of the teacher; and predicting the professional category of the teacher according to the problem solution records by using a fifth preset model.

Preferably, the courses include corresponding preset geographic areas, and the generating module 250 is further configured to obtain real-time locations of the users; determining the course to be a target course of the user to return to the user based on the relation between the real-time position and the preset geographic area.

In an embodiment of the present invention, there is also provided a computer system including:

determining the similarity between the user and the courses according to the course characteristics preferred by the user and the course characteristics corresponding to the courses of the target learning category;

FIG. 3 illustrates, among other things, an architecture of a computer system. As shown in fig. 3, may include a processor 1510, a video display adapter 1511, a disk drive 1512, an input/output interface 1513, a network interface 1514, and memory 1520. The processor 1510, video display adapter 1511, disk drive 1512, input/output interface 1513, network interface 1514, and memory 1520 may be communicatively coupled via a communication bus 1530.

The processor 1510 may be implemented by a general-purpose CPU (Central Processing Unit), a microprocessor, an Application Specific Integrated Circuit (ASIC), or one or more Integrated circuits, and is configured to execute related programs to implement the technical solution provided by the present Application.

The Memory 1520 may be implemented in the form of a ROM (Read Only Memory), a RAM (Random Access Memory), a static storage device, a dynamic storage device, or the like. The memory 1520 may store an operating system 1521 for controlling the operation of the computer system 1500, a Basic Input Output System (BIOS) for controlling low-level operations of the computer system 1500. In addition, a web browser 1523, a data storage management system 1524, an icon font processing system 1525, and the like can also be stored. The icon font processing system 1525 may be an application program that implements the operations of the foregoing steps in this embodiment of the application. In summary, when the technical solution provided by the present application is implemented by software or firmware, the relevant program codes are stored in the memory 1520 and called for execution by the processor 1510. The input/output interface 1513 is used for connecting an input/output module to realize information input and output. The i/o module may be configured as a component in a device (not shown) or may be external to the device to provide a corresponding function. The input devices may include a keyboard, a mouse, a touch screen, a microphone, various sensors, etc., and the output devices may include a display, a speaker, a vibrator, an indicator light, etc.

The network interface 1514 is used to connect a communication module (not shown) to enable the device to communicatively interact with other devices. The communication module can realize communication in a wired mode (such as USB, network cable and the like) and also can realize communication in a wireless mode (such as mobile network, WIFI, Bluetooth and the like).

The bus 1530 includes a path to transfer information between the various components of the device, such as the processor 1510, the video display adapter 1511, the disk drive 1512, the input/output interface 1513, the network interface 1514, and the memory 1520.

In addition, the computer system 1500 may also obtain information of specific extraction conditions from the virtual resource object extraction condition information database 1541 for performing condition judgment, and the like.

It should be noted that although the above devices only show the processor 1510, the video display adapter 1511, the disk drive 1512, the input/output interface 1513, the network interface 1514, the memory 1520, the bus 1530, etc., in a specific implementation, the devices may also include other components necessary for proper operation. Furthermore, it will be understood by those skilled in the art that the apparatus described above may also include only the components necessary to implement the solution of the present application, and not necessarily all of the components shown in the figures.

From the above description of the embodiments, it is clear to those skilled in the art that the present application can be implemented by software plus necessary general hardware platform. Based on such understanding, the technical solutions of the present application may be embodied in the form of a software product, which may be stored in a storage medium, such as a ROM/RAM, a magnetic disk, an optical disk, or the like, and includes several instructions for enabling a computer device (which may be a personal computer, a cloud server, or a network device) to execute the method according to the embodiments or some parts of the embodiments of the present application.

The embodiments in the present specification are described in a progressive manner, and the same and similar parts among the embodiments are referred to each other, and each embodiment focuses on the differences from the other embodiments. In particular, the system or system embodiments are substantially similar to the method embodiments and therefore are described in a relatively simple manner, and reference may be made to some of the descriptions of the method embodiments for related points. The above-described system and system embodiments are only illustrative, wherein the units described as separate parts may or may not be physically separate, and the parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of the present embodiment. One of ordinary skill in the art can understand and implement it without inventive effort.

The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents, improvements and the like that fall within the spirit and principle of the present invention are intended to be included therein.

Claims

1. A data processing method of a learning platform, the method comprising:

2. The method as claimed in claim 1, wherein the course learning record of the user includes a historical learning time of the user and a historical course score corresponding to the historical learning time, the method further comprising:

3. The method of claim 1 or 2, wherein determining the teacher's demonstration skill rating based on teacher video corresponding to the target learning category and determining the lesson to be the user's target lesson based on the teacher's demonstration skill rating, wherein the determining demonstration skill rating further comprises:

extracting image frames of the course video;

4. The method of claim 1 or 2, wherein determining the teacher's tone rating based on a teacher video corresponding to the target learning category and determining the lesson as the user's target lesson based on the teacher's tone rating, wherein the determining the tone rating further comprises:

extracting the audio of the course video;

5. The method as claimed in claim 1 or 2, wherein each of the lessons comprises a corresponding lesson video, the lesson features comprise keywords corresponding to the lesson, and the method further comprises: extracting the keywords, the extracting keywords further comprising:

6. The method according to claim 1 or 2, characterized in that the method further comprises:

acquiring a learning state image of the user when learning the target course;

7. The method as claimed in claim 1 or 2, wherein the determining the similarity between the user and the lesson according to the lesson characteristics of the user's preference and the lesson characteristics corresponding to the lesson of the target learning category further comprises:

8. The method of claim 1 or 2, wherein the user data comprises a search record of the user, the method comprising:

and generating a search record of the user according to the search keyword.

9. The method of claim 1 or 2, wherein the lesson features include a corresponding specialty category for the instructor, the method further comprising:

acquiring a problem solution record of a teacher;

10. The method of claim 1 or 2, wherein the lesson features include corresponding preset geographic areas, the method further comprising:

acquiring the real-time position of the user;