Online teaching method and system based on intelligent question and answer and data enhancement

Abstract

The invention discloses an online teaching method and system based on intelligent question and answer and data enhancement, and the method comprises the steps: S1, obtaining a text data set based on a question and answer mode according to an education material; s2, performing data enhancement on the text data set; s3, obtaining a question and answer knowledge prediction model according to the text data set after data enhancement; and S4, inputting the user question into the question and answer knowledge prediction model to obtain an optimal matching answer. By adopting the technical scheme of the invention, the questioning intention and the key entity of the user can be understood, and the direct answer required by the user is provided.

Description

Technical Field

[0001] The present invention belongs to the field of information processing technology, and in particular relates to an online teaching method and system based on intelligent question answering and data enhancement. Background Art

[0002] Globally, education is undergoing a digital transformation, driven by the popularity of the Internet, the development of online learning platforms, and the digitization of educational resources. The traditional education model faces increasing challenges due to its limitations, such as the inability to provide personalized learning paths and limited interaction between teachers and students. In order to address these issues, online education platforms need to introduce intelligent solutions to support a more adaptive and interactive learning experience.

[0003] With the rapid development of AI and NLP technologies, especially the successful application of deep learning models in natural language processing tasks, online education has ushered in a huge opportunity for innovation. The design goal of the intelligent question-answering system is to extract valuable information from massive educational data by combining AI technology to provide users with real-time and accurate help. Summary of the invention

[0004] The technical problem to be solved by the present invention is to provide an online teaching method and system based on intelligent question answering and data enhancement.

[0005] To achieve the above object, the present invention adopts the following technical solution:

[0006] An online teaching method based on intelligent question answering and data enhancement, including:

[0007] Step S1, obtaining a question-answer based text dataset according to educational materials;

[0008] Step S2, performing data enhancement on the text dataset;

[0009] Step S3, obtaining a question-answer knowledge prediction model based on the data-enhanced text dataset;

[0010] Step S4: input the user's question into the question-answering knowledge prediction model to obtain the best matching answer.

[0011] Preferably, in step S2, data enhancement is performed on the text dataset using a text enhancement model.

[0012] Preferably, in step S3, a neural network is trained according to the text data set after data augmentation to obtain a Q&A knowledge prediction model; wherein, the neural network includes: thirteen Conv2D convolutional layers, ten BN normalization layers, ten ReLU activation layers, a Concatenate splicing layer, six Add summation layers, a MaxPooling2D max pooling layer, a CBS module, an SPPF module, four LR-bneck modules, eight LRFEM modules, six LR-bneck modules, two LRConv modules, two Upsample modules, and four Concat modules.

[0013] Preferably, in step S1, teaching materials are collected from educational platforms, open courses, textbooks, academic papers, and monographs, and preprocessed and labeled to form a labeled Q&A-based text data set.

[0014] The present invention also provides an online teaching system based on intelligent Q&A and data augmentation, including:

[0015] An acquisition device for obtaining a Q&A-based text data set according to educational materials;

[0016] A processing device for performing data augmentation on the text data set;

[0017] A training device for obtaining a Q&A knowledge prediction model according to the text data set after data augmentation;

[0018] A prediction device for inputting a user's question into the Q&A knowledge prediction model to obtain the best-matched answer.

[0019] Preferably, the processing device performs data augmentation on the text data set through a text augmentation model.

[0020] Preferably, the training device trains a neural network according to the text data set after data augmentation to obtain a Q&A knowledge prediction model; wherein, the neural network includes: thirteen Conv2D convolutional layers, ten BN normalization layers, ten ReLU activation layers, a Concatenate splicing layer, six Add summation layers, a MaxPooling2D max pooling layer, a CBS module, an SPPF module, four LR-bneck modules, eight LRFEM modules, six LR-bneck modules, two LRConv modules, two Upsample modules, and four Concat modules.

[0021] Preferably, the acquisition device is used to collect teaching materials from educational platforms, open courses, textbooks, academic papers, and monographs, and preprocess and label them to form a labeled Q&A-based text data set.

[0022] The present invention can understand the user's question intention and key entities through a question-and-answer knowledge prediction model, and provide the direct answer required by the user. Description of the Drawings

[0023] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are only the embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained according to the provided drawings.

[0024] Figure 1 It is a flowchart of an online teaching method based on intelligent question answering and data enhancement according to an embodiment of the present invention. Detailed Embodiments

[0025] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of them. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.

[0026] To make the above objects, features, and advantages of the present invention more obvious and understandable, the present invention will be further described in detail below with reference to the drawings and specific embodiments.

[0027] Embodiment 1:

[0028] As Figure 1 shown, an online teaching method based on intelligent question answering and data enhancement according to an embodiment of the present invention includes:

[0029] Step S1: Obtain a question-and-answer-based text data set according to educational materials;

[0030] Step S2: Perform data enhancement on the text data set;

[0031] Step S3: Obtain a question-and-answer knowledge prediction model according to the text data set after data enhancement;

[0032] Step S4: Input the user's question into the question-and-answer knowledge prediction model to obtain the best-matched answer.

[0033] As an implementation manner of the embodiment of the present invention, teaching materials are collected from educational platforms, open courses, textbooks, academic papers, and monographs, and preprocessed and labeled to form a labeled question-and-answer-based text data set.

[0034] Further, the preprocessing and labeling are as follows:

[0035] Automatically detect and filter noise and redundant information in teaching materials using natural language processing techniques, and apply TF-IDF to evaluate the importance of vocabulary;

[0036] Use supervised NLP tools to automatically label samples, and verify the auxiliary automatic labeling results through the use of a crowdsourcing platform. Based on user feedback and verification results, form a labeled Q&A-based text dataset.

[0037] As an implementation manner of an embodiment of the present invention, in step S2, data augmentation is performed on the text dataset through a text augmentation model. Specifically, it includes:

[0038] Step S21: Obtain the text data of the text dataset;

[0039] Step S22: Input the text data into the first sub-model of the text augmentation model for the first augmentation process to obtain data-augmented text;

[0040] Step S23: Input the data-augmented text into the second sub-model of the text augmentation model for the second augmentation process to obtain augmented data.

[0041] Furthermore, the first augmentation process in the first diffusion sub-model includes:

[0042] Convert the text data into text vectors, and obtain the target noise vectors corresponding to the text data by continuously adding noise to each text vector multiple times;

[0043] Predict multiple noises added during the diffusion process, and use the target noise vectors corresponding to the text data to sequentially remove the predicted multiple noises to obtain the restored text vectors corresponding to the text data;

[0044] Convert the restored text vectors corresponding to the text data into text to obtain the data-augmented text corresponding to the text data.

[0045] Furthermore, the second augmentation process in the second sub-model includes:

[0046] Segment the data-augmented text to obtain a first sequence containing multiple words;

[0047] Randomly mask some words in the first sequence to obtain a second sequence;

[0048] Determine the label information of the first data, where the label information indicates that the data-augmented text is positive semantics or negative semantics;

[0049] Predict the masked part of the words in the first sequence according to the label information to obtain predicted words corresponding to the masked part of the words;

[0050] Concatenate the predicted word and the words in the second sequence to obtain enhanced data.

[0051] As an implementation manner of the embodiment of the present invention, in step S3, train a neural network according to the text data set after data enhancement to obtain a question and answer knowledge prediction model; wherein, the neural network includes: thirteen Conv2D convolutional layers, ten BN normalization layers, ten ReLU activation layers, a Concatenate layer, six Add summation layers, a MaxPooling2D max pooling layer, a CBS module, an SPPF module, four LR-bneck modules, eight LRFEM modules, six LR-bneck modules, two LRConv modules, two Upsample modules, and four Concat modules; wherein, the text data set after data enhancement is used as input data and first passes through the first Conv2D convolutional layer, and then passes through twelve Conv2D convolutional layers, ten BN normalization layers, ten ReLU activation layers, a Concatenate layer, six Add summation layers, a MaxPooling2D max pooling layer, a CBS module, an SPPF module, four LR-bneck modules, eight LRFEM modules, six LR-bneck modules, two LRConv modules, two Upsample modules, and four Concat modules according to the structure of the neural network to train the neural network.

[0052] The structure of the neural network is as follows:

[0053] The first Conv2D convolutional layer, the first BN normalization layer, and the first ReLU activation layer are connected in sequence; the output end of the first ReLU activation layer is respectively connected to the input ends of the second Conv2D convolutional layer, the third Conv2D convolutional layer, the fourth Conv2D convolutional layer, the fifth Conv2D convolutional layer, the sixth Conv2D convolutional layer, the seventh Conv2D convolutional layer, the eighth Conv2D convolutional layer, the ninth Conv2D convolutional layer, the tenth Conv2D convolutional layer, the eleventh Conv2D convolutional layer, the input end of the first Add summation layer, and the input end of the fourth Add summation layer; the output end of the first Add summation layer is connected to the input end of the fourth Add summation layer;

[0054] The second Conv2D convolutional layer, the second BN normalization layer, the second ReLU activation layer, the twelfth Conv2D convolutional layer, and the first Add summation layer are connected in sequence; the third Conv2D convolutional layer, the third BN normalization layer, the third ReLU activation layer, and the Concatenate splicing layer are connected in sequence; the fourth Conv2D convolutional layer, the fourth BN normalization layer, the fourth ReLU activation layer, and the Concatenate splicing layer are connected in sequence; the output end of the Concatenate splicing layer is connected to the input end of the fourth Add summation layer;

[0055] The fifth Conv2D convolutional layer, the fifth BN normalization layer, the fifth ReLU activation layer, and the second Add summation layer are connected in sequence; the sixth Conv2D convolutional layer, the sixth BN normalization layer, the sixth ReLU activation layer, and the second Add summation layer are connected in sequence; the seventh Conv2D convolutional layer, the seventh BN normalization layer, the seventh ReLU activation layer, and the second Add summation layer are connected in sequence;

[0056] The output end of the second Add summation layer is connected to the input end of the eleventh Conv2D convolutional layer; the output end of the eleventh Conv2D convolutional layer is connected to the input end of the third Add summation layer; the output end of the third Add summation layer is connected to the input end of the fourth Add summation layer; the output end of the fourth Add summation layer is connected to the input end of the MaxPooling2D max pooling layer;

[0057] The eighth Conv2D convolutional layer, the eighth BN normalization layer, the eighth ReLU activation layer, and the fifth Add summation layer are connected in sequence; the ninth Conv2D convolutional layer, the ninth BN normalization layer, the ninth ReLU activation layer, and the fifth Add summation layer are connected in sequence; the tenth Conv2D convolutional layer, the tenth BN normalization layer, the tenth ReLU activation layer, and the fifth Add summation layer are connected in sequence; the fifth Add summation layer, the thirteenth Conv2D convolutional layer, and the sixth Add summation layer are connected in sequence; the output end of the sixth Add summation layer is connected to the input end of the fourth Add summation layer;

[0058] The output end of the MaxPooling2D max pooling layer, the CBS module, the first LR-bneck module, the first LRFEM module, the second LR-bneck module, the second LRFEM module, the third LR-bneck module, the third LRFEM module, the fourth LR-bneck module, the fourth LRFEM module, and the SPPF module are connected in sequence; the first LRConv module, the first Upsample module, the first Concat module, the fifth LRFEM module, the second LRConv module, the second Upsample module, the second Concat module, and the sixth LRFEM module are connected in sequence; the second LRConv module is connected to the third Concat module, the third Concat module is connected to the seventh LRFEM module and the fifth LR-bneck module, the fifth LR-bneck module is connected to the sixth LRFEM module, and the seventh LRFEM module is connected to the sixth LR-bneck module, the fourth Concat module, and the eighth LRFEM module in sequence; the first LRConv module is connected to the SPPF module and the fourth Concat module, the second LRFEM module is connected to the second Concat module, and the third LRFEM module is connected to the first Concat module.

[0059] Among them, the kernels of the first Conv2D convolutional layer and the second Conv2D convolutional layer are both 3×3, the kernel of the third Conv2D convolutional layer is 3×1, the kernel of the fourth Conv2D convolutional layer is 1×3, the kernel of the fifth Conv2D convolutional layer is 5×5, the kernel of the sixth Conv2D convolutional layer is 5×1, the kernel of the seventh Conv2D convolutional layer is 1×5, the kernel of the eighth Conv2D convolutional layer is 7×7, the kernel of the ninth Conv2D convolutional layer is 7×1, the kernel of the tenth Conv2D convolutional layer is 1×7, and the kernels of the eleventh Conv2D convolutional layer, the twelfth Conv2D convolutional layer, and the thirteenth Conv2D convolutional layer are all 1×1.

[0060] As an implementation manner of the embodiment of the present invention, in step S4, the user question is input into the question and answer knowledge prediction model to obtain the best-matched answer.

[0061] For example, for the online teaching of a training course on "The Development Trends and Innovations in the Enterprise Training Market", the course content includes: the current situation analysis of the enterprise training market, the development trends of the enterprise training market, the innovative methods and strategies of enterprise training, case analysis and discussion, and practical operation and sharing; among them, the questions and answers of the teacher-student teaching interaction are realized through the question and answer knowledge prediction model in the teaching interaction of the course content of "The Development Trends and Innovations in the Enterprise Training Market", specifically:

[0062] Student's question:

[0063] "Teacher, what do you think are the main trends in the future corporate training market?"

[0064] The teacher replied:

[0065] "The future corporate training market will place more emphasis on personalized and customized training programs. With the development of technologies such as artificial intelligence and big data, enterprises can more accurately identify the learning needs of employees and provide targeted training content. Blended learning will also become mainstream, combining the advantages of online and offline to offer more flexible learning methods."

[0066] The student asked:

[0067] "In corporate training, how can innovative methods be effectively introduced?"

[0068] The teacher replied:

[0069] "The key to introducing innovative methods lies in understanding new technologies and tools and combining them with the specific needs of the enterprise. For example, using virtual reality (VR) for skills training or leveraging gamification to improve employee engagement. Establishing a cultural environment that supports innovation is also very important, encouraging employees to try new methods and learn from them."

[0070] The student asked:

[0071] "In the future training market, which emerging technologies are likely to become mainstream?"

[0072] The teacher replied:

[0073] "Based on current trend analysis, technologies such as virtual reality (VR), augmented reality (AR), artificial intelligence (AI), and big data analytics are expected to play important roles in the future corporate training market."

[0074] Example 2:

[0075] This embodiment of the present invention also provides an online teaching system based on intelligent Q&A and data augmentation, including:

[0076] An acquisition device for obtaining a Q&A-based text data set according to educational materials;

[0077] A processing device for performing data augmentation on the text data set;

[0078] A training device for obtaining a Q&A knowledge prediction model according to the text data set after data augmentation;

[0079] A prediction device for inputting a user question into the Q&A knowledge prediction model to obtain the best-matched answer.

[0080] As an implementation manner of an embodiment of the present invention, the processing device performs data augmentation on the text data set through a text augmentation model.

[0081] As an implementation manner of an embodiment of the present invention, the training device trains a neural network according to the text data set after data augmentation to obtain a question-and-answer knowledge prediction model; wherein, the neural network includes: thirteen Conv2D convolutional layers, ten BN normalization layers, ten ReLU activation layers, a Concatenate splicing layer, six Add summation layers, a MaxPooling2D max pooling layer, a CBS module, an SPPF module, four LR-bneck modules, eight LRFEM modules, six LR-bneck modules, two LRConv modules, two Upsample modules, and four Concat modules.

[0082] As an implementation manner of an embodiment of the present invention, an acquisition device is configured to collect teaching materials from educational platforms, open courses, textbooks, academic papers, and monographs, and perform preprocessing and annotation to form a labeled question-and-answer-based text data set.

[0083] The above-described embodiments are only descriptions of the preferred embodiments of the present invention, and do not limit the scope of the present invention. Without departing from the design spirit of the present invention, various deformations and improvements made by those of ordinary skill in the art to the technical solution of the present invention shall fall within the protection scope determined by the claims of the present invention.

Claims

1. An online teaching method based on intelligent question answering and data enhancement, characterized in that: include: Step S1, obtaining a question-answer based text dataset according to educational materials; Step S2, performing data enhancement on the text data set; Step S3, obtaining a question-answer knowledge prediction model based on the data-enhanced text dataset; Step S4: input the user's question into the question-answering knowledge prediction model to obtain the best matching answer.

2. The online teaching method based on intelligent question answering and data enhancement as claimed in claim 1, characterized in that: In step S2, the text dataset is enhanced by using a text enhancement model.

3. The online teaching method based on intelligent question answering and data enhancement as claimed in claim 2, characterized in that: In step S3, a neural network is trained according to the data-enhanced text data set to obtain a question-answer knowledge prediction model; wherein the neural network includes: thirteen Conv2D convolution layers, ten BN normalization layers, ten ReLU activation layers, Concatenate splicing layers, six Add summation layers, MaxPooling2D maximum pooling layers, CBS modules, SPPF modules, four LR-bneck modules, eight LRFEM modules, six LR-bneck modules, two LRConv modules, two Upsample modules, and four Concat modules.

4. The online teaching method based on intelligent question answering and data enhancement as claimed in claim 3, characterized in that: In step S1, teaching materials are collected from educational platforms, open courses, textbooks, academic papers, and monographs, and preprocessed and annotated to form a labeled question-answering-based text dataset.

5. An online teaching system based on intelligent question answering and data enhancement, characterized in that: include: An acquisition device, for obtaining a text dataset based on a question-answer format according to educational materials; A processing device, performs data enhancement on the text dataset; A training device obtains a question-answering knowledge prediction model based on the data-enhanced text dataset; The prediction device inputs the user's question into the question-answering knowledge prediction model to obtain the best matching answer.

6. The online teaching system based on intelligent question answering and data enhancement as claimed in claim 5, characterized in that: The processing device performs data enhancement on the text dataset through a text enhancement model.

7. The online teaching system based on intelligent question answering and data enhancement as claimed in claim 6, characterized in that: The training device trains a neural network according to the data-enhanced text data set to obtain a question-answer knowledge prediction model; wherein the neural network includes: thirteen Conv2D convolution layers, ten BN normalization layers, ten ReLU activation layers, Concatenate splicing layers, six Add summation layers, MaxPooling2D maximum pooling layer, CBS module, SPPF module, four LR-bneck modules, eight LRFEM modules, six LR-bneck modules, two LRConv modules, two Upsample modules, and four Concat modules.

8. The online teaching system based on intelligent question answering and data enhancement as claimed in claim 7, characterized in that: The acquisition device is used to collect teaching materials from educational platforms, open courses, textbooks, academic papers, and monographs, and preprocess and annotate them to form a labeled question-answering text dataset.

Images