Video education resource query method and system based on natural language processing

Abstract

The present invention discloses a video education resource query method and system based on natural language processing, which relates to the field of video education technology, in order to solve the problem that user retrieval data does not match the presented video. The present invention determines the final category of the video according to the attribute category of the keyword, which can achieve fast and accurate classification, and uniformly classifies educational resource videos of the same category, which can achieve centralized management and optimal configuration of resources, facilitate subsequent retrieval and use, convert standard language data and educational resource videos into curve data, so that the two have the same data expression form, which is convenient for subsequent comparison and analysis, and compares by overlapping curves, which can more intuitively and accurately judge the similarity between the two, and by formulating a scheme presentation module to perform threshold level classification, it can provide users with differentiated presentation schemes according to different similarities.

Description

Technical Field

[0001] The present invention relates to the field of video education technology, and in particular to a video education resource query method and system based on natural language processing. Background Art

[0002] Video educational resources refer to educational resources that convey knowledge and information through video.

[0003] The Chinese patent with publication number CN113722539A discloses a video classification method, device, electronic device and storage medium. The target cluster corresponding to the image frame vector of the image frame in the target video can be determined to obtain the to-be-recognized text corresponding to the target video, so that when classifying the target video, the image frame in the target video is converted into the text content in the to-be-recognized text, thereby reducing the computational complexity, shortening the processing time, and reducing the requirements for the processing capability of the device. Although the above patent solves the problem of video classification, the following problems still exist in actual operation:

[0004] 1. Failure to perform similarity matching between user search terms and videos in video education resources, resulting in reduced accuracy in user search data presentation.

[0005] 2. The videos on the platform are not classified by video attributes, which results in a mismatch between the videos on the platform and the user's search data.

[0006] 3. The user's retrieval data is not further processed and optimized, so the key information of the retrieval data cannot be accurately extracted, resulting in subsequent video matching errors. Summary of the invention

[0007] The purpose of the present invention is to provide a video education resource query method and system based on natural language processing. The final category of the video can be determined according to the attribute category of the keyword, so as to achieve fast and accurate classification. Educational resource videos of the same category are uniformly classified, so as to achieve centralized management and optimal configuration of resources, and facilitate subsequent retrieval and use. Standard language data and educational resource videos are converted into curve data so that the two have the same data expression form, which is convenient for subsequent comparison and analysis. By comparing them in the form of curve overlap, the similarity between the two can be judged more intuitively and accurately. By formulating a scheme presentation module for threshold level classification, differentiated presentation schemes can be provided to users according to different similarities, which can solve the problems in the prior art.

[0008] To achieve the above object, the present invention provides the following technical solutions:

[0009] The video education resource query method based on natural language processing includes the following steps:

[0010] S1: Receiving search data statistics: The user inputs the search term into the search platform, collects the search platform and the search term data, and stores the search platform and the search term data independently;

[0011] S2: Retrieval data analysis: Perform natural language processing on the user's retrieval data, where natural language processing includes semantic information analysis and key information analysis of the retrieval terms, classify the retrieval terms after natural language processing into attribute categories, and mark them as standard language data;

[0012] S3: Retrieval platform matching: Confirm the video resource library in the retrieval platform, analyze the videos in the video resource library by language category, and classify the videos in the video resource library by category according to the analysis results;

[0013] S4: Video resource presentation: Import the standard language data into the video resource library for category matching, confirm the similarity of the standard language data matching, import the similarity threshold into the presentation rule to formulate a presentation plan, and present the formulated plan to the display terminal.

[0014] Preferably, the search platform and search term data are collected in S1, including:

[0015] The user first logs in to the search platform and enters the search data in the search engine of the search platform;

[0016] Confirm the IP address of the search platform where the user logs in;

[0017] Import the retrieval platform IP address into the address database for security verification;

[0018] Among them, security verification is to verify the IP address of the retrieval platform with the IP in the address database;

[0019] When the IP address of the retrieval platform is verified to be qualified, the retrieval platform is a safe platform;

[0020] When the IP address verification of the search platform fails, the search platform is considered an abnormal platform and will not accept search terms.

[0021] The security platform and the search terms in the security platform search engine are stored separately.

[0022] Preferably, natural language processing is performed on the user's search data in S2, including:

[0023] Before retrieving data for natural language processing, the retrieved data is preprocessed;

[0024] Among them, the search data is first cleaned, and the data cleaning is to eliminate abnormal values, missing values ​​and duplicate values ​​in the search data;

[0025] The search data after data cleaning is converted, wherein the data conversion includes conversion of numerical data and text data in the search data, conversion of categorical variables and continuous variables, and normalization of features;

[0026] Normalizing the search data after data conversion, wherein data normalization is to adjust the dimension and range of the search data to be consistent;

[0027] After data normalization, the retrieval data with data preprocessing completed is obtained.

[0028] Preferably, the semantic information analysis and key information analysis of the search terms in S2 further includes:

[0029] The retrieval data after data preprocessing is analyzed for semantic information and key information in turn;

[0030] Among them, the search data is first subjected to semantic information analysis, and the semantic information analysis is to extract feature vectors from the search data through the bag-of-words model;

[0031] The extracted feature vectors are used for model training through a computer learning model;

[0032] Perform inference on the retrieval data after model training, and obtain the semantic information of the retrieval data after inference;

[0033] In the inference process, the feature vector of the retrieved data is predicted by the model, and the semantic information of the retrieved data is obtained after the prediction. The semantic information includes sentiment analysis, topic classification, and question-answering system;

[0034] Finally, the semantic information of the retrieved data is obtained;

[0035] Then, the search data is subjected to key information extraction. The key information extraction uses the bag-of-words model to extract keyword-related features in the search data, where the keyword-related features are text, images, and audio.

[0036] Import the extracted keyword information into the semantic information for keyword selection;

[0037] The selected keywords are used as the key information for retrieving data.

[0038] 6. Preferably, the retrieval data after model training is inferred to obtain semantic information of the retrieval data after inference, including:

[0039] The retrieval data after model training is processed in a data planning manner according to the target reasoning language conversion strategy to obtain the target retrieval data;

[0040] Based on the semantic analysis dimension, determining a marking rule for the target retrieval data, marking the target retrieval data according to the marking rule, and obtaining target marked data;

[0041] Based on the marking features of the target marked data, feature vectors are extracted from the target marked data to obtain a feature vector group corresponding to each semantic analysis dimension, and a dimension matrix is ​​constructed based on the feature vector group;

[0042] Based on the matrix characteristics of the dimension matrix, determine the reasoning scheme for different semantic analysis dimensions, and process the initial dimension analysis reasoning model based on the reasoning scheme to obtain the target dimension analysis reasoning model;

[0043] Input the dimension matrix into the corresponding target dimension analysis and reasoning model, and output the initial reasoning information and the associated data and supplementary data related to the dimension matrix;

[0044] Obtaining initial semantic information of the retrieved data based on the initial reasoning information;

[0045] Determining whether the correlation between the initial semantic information and the associated data and the supplementary data is greater than a preset correlation;

[0046] If so, determining that the initial semantic information is the final semantic information of the retrieved data;

[0047] Otherwise, based on the difference between the correlation between the initial semantic information and the associated data and supplementary data and the preset correlation, the adjustment weight is determined, and based on the adjustment weight, combined with the reference semantic information of the associated data and supplementary data, the initial semantic information is adjusted to obtain the final semantic information of the retrieved data.

[0048] Preferably, the search terms processed by natural language processing in S2 are classified into attribute categories, including:

[0049] Determine the attribute category of the search data based on the key information, where the attribute category is the subject of the search data, including science, art, history, politics, geography, language, etc.;

[0050] Import the key information into the subject database for attribute category confirmation, and obtain the standard language data after the attribute category confirmation of the retrieved data is completed.

[0051] Preferably, the language category analysis is performed on the videos in the video resource library in S3, and the videos in the video resource library are classified according to the category according to the analysis result, including:

[0052] Confirm each educational resource video in the video resource library;

[0053] Then, the language and text of each educational resource video are obtained;

[0054] Extract keywords from the language and text of educational resource videos;

[0055] Performing attribute category judgment on the keywords extracted from the language and text of the educational resource video, wherein the attribute category of the keywords extracted from the language and text in the educational resource video is consistent with the attribute category of the search term;

[0056] Determine the final category of the educational resource video according to the keyword attribute category extracted from the language and text in the educational resource video;

[0057] And the educational resource videos of the same category are uniformly classified.

[0058] A video education resource query method based on natural language processing, according to the video education resource query method based on natural language processing, comprising:

[0059] Similarity confirmation module, used for:

[0060] Match the attribute categories of the standard language data with the attribute categories of the educational resource videos in the video resource library;

[0061] Perform curve data conversion on standard language data and educational resource videos with the same attribute category;

[0062] Overlapping the curve data of the standard language data with the curve data of the educational resource video;

[0063] The similarity between the standard language data and the educational resource video is determined based on the range of the overlapping area when the curves overlap;

[0064] Among them, the more overlapping areas there are, the higher the similarity between the standard language data and the educational resource video.

[0065] Preferably, the similarity confirmation module includes:

[0066] An overlapping unit, used for overlapping the curve data of the standard language data with the curve data of the educational resource video to obtain an overlapping area;

[0067] An area calculation unit, used for calculating the area S of the overlapping area based on the curve characteristics of the standard language data and the curve characteristics of the educational resource video in the overlapping area according to the following formula;

[0068]

[0069] Where n represents the number of intersections between the curve data of the standard language data and the curve data of the educational resource video, S 0i Indicates S aiand S bi The smaller of ai represents the area corresponding to the curve feature of the standard language data corresponding to the i-th intersection, S bi represents the area corresponding to the curve feature of the educational resource video corresponding to the i-th intersection, and ΔSr represents the reference standard area difference;

[0070] A similarity calculation unit, for calculating the similarity P between the standard language data and the educational resource video based on the area of ​​the overlapping area and the curve data of the standard language data according to the following formula;

[0071]

[0072] Among them, S V It represents the similarity reference area corresponding to the category attributes of the standard language data and the educational resource video. e represents a natural constant with a value of 2.72. σ The matching weight of the curve data representing the standard language data and the curve data of the educational resource video and the semantic information is (0, 1).

[0073] Preferably, it also includes:

[0074] Develop a program presentation module to:

[0075] Classify the similarity between the standard language data and the educational resource videos by threshold level;

[0076] Among them, the threshold levels are divided into primary level, secondary level and tertiary level;

[0077] The presentation plan is formulated according to the threshold level classification of similarity;

[0078] The formulated plans are presented on the display terminal in sequence according to the threshold levels.

[0079] Compared with the prior art, the present invention has the following beneficial effects:

[0080] 1. The video education resource query method and system based on natural language processing provided by the present invention can improve processing efficiency and accuracy by preprocessing data, provide a high-quality data basis for subsequent natural language processing tasks, and perform semantic information and key information analysis on the retrieval data after data preprocessing, so as to provide richer, more accurate and diversified retrieval results, and improve the retrieval accuracy and user satisfaction.

[0081] 2. The video education resource query method and system based on natural language processing provided by the present invention can determine the final category of the video according to the attribute category of the keyword, which can achieve fast and accurate classification and improve classification efficiency. The educational resource videos of the same category are uniformly classified, which can realize centralized management and optimal configuration of resources and facilitate subsequent retrieval and use.

[0082] 3. The video education resource query method and system based on natural language processing provided by the present invention convert standard language data and education resource video into curve data, so that the two have the same data expression form, which is convenient for subsequent comparison and analysis. By comparing by overlapping curves, the similarity between the two can be judged more intuitively and accurately. By formulating a scheme presentation module for threshold level classification, differentiated presentation schemes can be provided to users according to different similarities. BRIEF DESCRIPTION OF THE DRAWINGS

[0083] Figure 1 A schematic diagram of the steps of the video education resource query method based on natural language processing of the present invention;

[0084] Figure 2 This is a flow diagram of the video education resource query system based on natural language processing of the present invention. DETAILED DESCRIPTION

[0085] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0086] In order to solve the problem in the prior art that the user's search data is not further processed and optimized, so that the key information of the search data cannot be accurately extracted, resulting in the subsequent video matching error, please refer to Figure 1 and Figure 2 , this embodiment provides the following technical solutions:

[0087] The video education resource query method based on natural language processing includes the following steps:

[0088] S1: Receiving search data statistics: The user inputs the search term into the search platform, collects the search platform and the search term data, and stores the search platform and the search term data independently;

[0089] Among them, users only search on platforms with qualified IP address verification, which can ensure the stability and reliability of the search process and enhance the user experience;

[0090] S2: Retrieval data analysis: Perform natural language processing on the user's retrieval data, where natural language processing includes semantic information analysis and key information analysis of the retrieval terms, classify the retrieval terms after natural language processing into attribute categories, and mark them as standard language data;

[0091] Among them, by judging the attribute category, the search data can be classified and managed. Users can select the corresponding attribute category according to their needs to query and obtain information that better meets their needs. The establishment of a subject library helps to focus on the search data of specific topics, making the relevant data more centralized and easier to find;

[0092] S3: Retrieval platform matching: Confirm the video resource library in the retrieval platform, analyze the videos in the video resource library by language category, and classify the videos in the video resource library by category according to the analysis results;

[0093] Among them, by comparing the consistency of the attribute categories of keywords and the attribute categories of search terms, the accuracy and reliability of classification can be improved. By judging the final category of the video according to the attribute categories of keywords, fast and accurate classification can be achieved, thereby improving classification efficiency.

[0094] S4: Video resource presentation: importing the standard language data into the video resource library for category matching, confirming the similarity of the standard language data matching, importing the similarity threshold into the presentation rule to formulate a presentation plan, and presenting the formulated plan to the display terminal;

[0095] Among them, by comparing by means of curve overlap, the similarity between the two can be judged more intuitively and accurately. The more overlapping areas there are, the higher the similarity between the two is, so that differentiated presentation solutions can be provided to users according to the similarity threshold level between the standard language data and the educational resource video.

[0096] The search platform and search term data in S1 are collected, including:

[0097] The user first logs in to the search platform and enters the search data in the search engine of the search platform;

[0098] Confirm the IP address of the search platform where the user logs in;

[0099] Import the retrieval platform IP address into the address database for security verification;

[0100] Among them, security verification is to verify the IP address of the retrieval platform with the IP in the address database;

[0101] When the IP address of the retrieval platform is verified to be qualified, the retrieval platform is a safe platform;

[0102] When the IP address verification of the search platform fails, the search platform is considered an abnormal platform and will not accept search terms.

[0103] The security platform and the search terms in the security platform search engine are stored separately.

[0104] Specifically, IP address verification can effectively prevent illegal access and potential security threats, improve the security of the retrieval platform, and users only search on platforms with qualified IP address verification, which can ensure the stability and reliability of the retrieval process and enhance the user experience. All security platforms and search terms are stored for easy subsequent management and analysis, which helps to understand user needs and behaviors, and adjust IP address verification rules and standards as needed to adapt to different security requirements and user groups.

[0105] Perform natural language processing on the user's search data in S2, including:

[0106] Before retrieving data for natural language processing, the retrieved data is preprocessed;

[0107] Among them, the search data is first cleaned, and the data cleaning is to eliminate abnormal values, missing values ​​and duplicate values ​​in the search data;

[0108] The search data after data cleaning is converted, wherein the data conversion includes conversion of numerical data and text data in the search data, conversion of categorical variables and continuous variables, and normalization of features;

[0109] Normalizing the search data after data conversion, wherein data normalization is to adjust the dimension and range of the search data to be consistent;

[0110] After data normalization, the retrieval data with data preprocessing completed is obtained.

[0111] Specifically, data preprocessing is an indispensable part of natural language processing. It can improve processing efficiency and accuracy and provide a high-quality data foundation for subsequent natural language processing tasks. First, the retrieval data is cleaned to eliminate outliers, missing values, and duplicate values. This step can ensure the accuracy and reliability of the data and avoid interference with subsequent processing due to data quality issues. After completing data cleaning, data conversion is performed. This includes the conversion of numerical data and text data, the conversion of categorical variables and continuous variables, and the normalization of features. Data conversion can adapt data to specific processing requirements, improve the efficiency and accuracy of subsequent natural language processing, and normalize the converted data to make the dimension and range of the data consistent. Normalization can eliminate the dimensional differences between different features, making each feature on the same scale, which helps to improve the stability and generalization ability of the model. Through data cleaning and conversion, the accuracy and consistency of the data can be ensured, and the impact of errors and anomalies on subsequent processing can be reduced. Appropriate data conversion and normalization can improve the efficiency of natural language processing, allowing the model to learn and process data faster. Normalized data helps to improve the generalization ability of the model, enabling it to better adapt to different data distributions and scenarios.

[0112] The retrieval data after data preprocessing is analyzed for semantic information and key information in turn;

[0113] Among them, the search data is first subjected to semantic information analysis, and the semantic information analysis is to extract feature vectors from the search data through the bag-of-words model;

[0114] The extracted feature vectors are used for model training through a computer learning model;

[0115] Perform inference on the retrieval data after model training, and obtain the semantic information of the retrieval data after inference;

[0116] In the inference process, the feature vector of the retrieved data is predicted by the model, and the semantic information of the retrieved data is obtained after the prediction. The semantic information includes sentiment analysis, topic classification, and question-answering system;

[0117] Finally, the semantic information of the retrieved data is obtained;

[0118] Then, the search data is subjected to key information extraction. The key information extraction uses the bag-of-words model to extract keyword-related features in the search data, where the keyword-related features are text, images, and audio.

[0119] Import the extracted keyword information into the semantic information for keyword selection;

[0120] The selected keywords are used as the key information for retrieving data.

[0121] Specifically, performing semantic information and key information analysis on the retrieval data after data preprocessing can provide richer, more accurate and diverse retrieval results, improve the accuracy of retrieval and user satisfaction. Through semantic information analysis, semantic information such as sentiment analysis, topic classification, and question-answering system can be extracted from the retrieval data. This information helps to deeply understand the user's query intention and provide more accurate and personalized retrieval results. Through model training and reasoning, the user's query intention can be understood more accurately, thereby improving the accuracy of retrieval. This helps users find the information they need faster and improves user satisfaction. Key information extraction is not limited to text, but can also include other types of information such as images and audio. This helps to provide more diverse retrieval results and meet different types of user needs. Through keyword selection, the keywords most relevant to the user's query intention can be screened out as key information. This helps to improve the focus and relevance of retrieval results.

[0122] In one embodiment, the retrieval data after model training is inferred to obtain semantic information of the retrieval data after inference, including:

[0123] The retrieval data after model training is processed in a data planning manner according to the target reasoning language conversion strategy to obtain the target retrieval data;

[0124] Based on the semantic analysis dimension, determining a marking rule for the target retrieval data, marking the target retrieval data according to the marking rule, and obtaining target marked data;

[0125] Based on the marking features of the target marked data, feature vectors are extracted from the target marked data to obtain a feature vector group corresponding to each semantic analysis dimension, and a dimension matrix is ​​constructed based on the feature vector group;

[0126] Based on the matrix characteristics of the dimension matrix, determine the reasoning scheme for different semantic analysis dimensions, and process the initial dimension analysis reasoning model based on the reasoning scheme to obtain the target dimension analysis reasoning model;

[0127] Input the dimension matrix into the corresponding target dimension analysis and reasoning model, and output the initial reasoning information and the associated data and supplementary data related to the dimension matrix;

[0128] Obtaining initial semantic information of the retrieved data based on the initial reasoning information;

[0129] Determining whether the correlation between the initial semantic information and the associated data and the supplementary data is greater than a preset correlation;

[0130] If so, determining that the initial semantic information is the final semantic information of the retrieved data;

[0131] Otherwise, based on the difference between the correlation between the initial semantic information and the associated data and supplementary data and the preset correlation, the adjustment weight is determined, and based on the adjustment weight, combined with the reference semantic information of the associated data and supplementary data, the initial semantic information is adjusted to obtain the final semantic information of the retrieved data.

[0132] In this embodiment, the retrieval data after model training is processed in a data planning manner according to the target reasoning language conversion strategy. The purpose of obtaining the target retrieval data is to provide a high-quality initial data basis for the reasoning of the retrieval data.

[0133] In this embodiment, the tagging rules are determined according to the features of the semantic analysis dimension.

[0134] In this embodiment, the semantic analysis dimensions include sentiment analysis, topic classification, and question-answering system.

[0135] In this embodiment, the dimension matrix is ​​used to represent the semantic analysis dimension retrieval data features.

[0136] The beneficial effects of the above design scheme are: analyzing the semantic information and key information of the retrieval data after data preprocessing can provide richer, more accurate and diversified retrieval results, improve the accuracy of retrieval and user satisfaction, and extract semantic information such as sentiment analysis, topic classification, and question-answering system in the retrieval data through semantic information analysis. This information helps to deeply understand the user's query intention and provide more accurate and personalized retrieval results. Through model training and reasoning, the user's query intention can be understood more accurately, thereby improving the accuracy of retrieval. By reasoning the retrieval data from the semantic analysis dimension, the initial reasoning information and the associated data and supplementary data related to the dimension matrix are output, and the initial reasoning information is verified and adjusted based on the associated data and supplementary data, so as to ensure the accuracy of the semantic information of the final retrieval data.

[0137] The search terms that have been processed by natural language in S2 are classified into attribute categories, including:

[0138] Determine the attribute category of the search data based on the key information, where the attribute category is the subject of the search data, including science, art, history, politics, geography, language, etc.;

[0139] Import the key information into the subject database for attribute category confirmation, and obtain the standard language data after the attribute category confirmation of the retrieved data is completed.

[0140] Specifically, the attribute category of the search data is determined through key information, and the search data is classified according to the attribute category, which further improves the management and use efficiency of the search data and provides users with more accurate and professional information. The search data can be classified and managed by judging the attribute category. This helps to improve the management and use efficiency of data, making the search data more orderly and easy to query. Classifying the search data according to the attribute category can provide users with more professional and accurate information. Users can select the corresponding attribute category to query according to their needs and obtain information that better meets their needs. The establishment of a subject library helps to focus the search data on specific topics, making the relevant data more concentrated and easier to find. This helps to improve the accuracy and efficiency of retrieval.

[0141] In order to solve the problem in the prior art that the videos on the platform are not classified by video attributes, which leads to the inability to match the videos on the platform with the user's search data, please refer to Figure 1 and Figure 2 , this embodiment provides the following technical solutions:

[0142] Perform language category analysis on the videos in the video resource library in S3, and classify the videos in the video resource library according to the analysis results, including:

[0143] Confirm each educational resource video in the video resource library;

[0144] Then, the language and text of each educational resource video are obtained;

[0145] Extract keywords from the language and text of educational resource videos;

[0146] Performing attribute category judgment on the keywords extracted from the language and text of the educational resource video, wherein the attribute category of the keywords extracted from the language and text in the educational resource video is consistent with the attribute category of the search term;

[0147] Determine the final category of the educational resource video according to the keyword attribute category extracted from the language and text in the educational resource video;

[0148] And the educational resource videos of the same category are uniformly classified.

[0149] Specifically, by confirming each educational resource video, the integrity and accuracy of the video resources can be ensured, and omissions or incorrect classifications can be avoided. By extracting the language and text of each video, key information can be extracted from it to provide a basis for subsequent classification. By comparing the consistency of the attribute categories of keywords and the attribute categories of search terms, the accuracy and reliability of classification can be improved. By judging the final category of the video based on the attribute categories of keywords, fast and accurate classification can be achieved, and classification efficiency can be improved. Unified classification of educational resource videos of the same category can achieve centralized management and optimal allocation of resources, which is convenient for subsequent retrieval and use.

[0150] In order to solve the problem in the prior art that the similarity matching between the user's search terms and the videos in the video education resources is not performed, thereby reducing the accuracy of the user's search data presentation, please refer to Figure 1 and Figure 2 , this embodiment provides the following technical solutions

[0151] A video education resource query method based on natural language processing, according to the video education resource query method based on natural language processing, comprising:

[0152] Similarity confirmation module, used for:

[0153] Match the attribute categories of the standard language data with the attribute categories of the educational resource videos in the video resource library;

[0154] Perform curve data conversion on standard language data and educational resource videos with the same attribute category;

[0155] Overlapping the curve data of the standard language data with the curve data of the educational resource video;

[0156] The similarity between the standard language data and the educational resource video is determined based on the range of the overlapping area when the curves overlap;

[0157] Among them, the more overlapping areas there are, the higher the similarity between the standard language data and the educational resource video.

[0158] In one embodiment, the similarity confirmation module includes:

[0159] An overlapping unit, used for overlapping the curve data of the standard language data with the curve data of the educational resource video to obtain an overlapping area;

[0160] An area calculation unit, used for calculating the area S of the overlapping area based on the curve characteristics of the standard language data and the curve characteristics of the educational resource video in the overlapping area according to the following formula;

[0161]

[0162] Where n represents the number of intersections between the curve data of the standard language data and the curve data of the educational resource video, S 0i Indicates S ai and S bi The smaller of ai represents the area corresponding to the curve feature of the standard language data corresponding to the i-th intersection, S bi represents the area corresponding to the curve feature of the educational resource video corresponding to the i-th intersection, and ΔSr represents the reference standard area difference;

[0163] A similarity calculation unit, for calculating the similarity P between the standard language data and the educational resource video based on the area of ​​the overlapping area and the curve data of the standard language data according to the following formula;

[0164]

[0165] Among them, S V It represents the similarity reference area corresponding to the category attributes of the standard language data and the educational resource video, e represents a natural constant, and its value is 2.72, σ represents the matching weight of the curve data of the standard language data and the curve data of the educational resource video and the semantic information, and its value is (0, 1).

[0166] In this embodiment, the matching weights of the curve data of the standard language data and the curve data of the educational resource video and the semantic information are used to represent the proportional matching feature between the curve data and the semantic information.

[0167] The beneficial effect of the above design is that by comparing by means of curve overlap, the similarity between the two can be judged more intuitively and accurately. The more overlapping areas there are, the higher the similarity between the two. By uniformly processing and comparing the standard language data and the educational resource videos with the same attribute category, the efficiency of comparison can be improved, so that the similarity between the standard language data and the educational resource videos can be accurately and efficiently evaluated, providing a reliable basis for subsequent data processing and analysis, and combining the curve data characteristics and the data attribute characteristics to finally obtain the similarity, ensuring the accuracy of the obtained similarity.

[0168] Develop a program presentation module to:

[0169] Classify the similarity between the standard language data and the educational resource videos by threshold level;

[0170] Among them, the threshold levels are divided into primary level, secondary level and tertiary level;

[0171] The presentation plan is formulated according to the threshold level classification of similarity;

[0172] The formulated plans are presented on the display terminal in sequence according to the threshold levels.

[0173] Specifically, the accuracy of the comparison can be ensured by matching the attribute categories through the similarity confirmation module. Only when the attribute category of the standard language data is the same as the attribute category of the educational resource video, the subsequent comparison is carried out, and the standard language data and the educational resource video are converted into curve data so that the two have the same data expression form, which is convenient for subsequent comparison and analysis. By comparing by overlapping curves, the similarity between the two can be judged more intuitively and accurately. The more overlapping areas, the higher the similarity between the two. By uniformly processing and comparing the standard language data and the educational resource video with the same attribute category, the efficiency of the comparison can be improved, so that the similarity between the standard language data and the educational resource video can be accurately and efficiently evaluated, providing a reliable basis for subsequent data processing and analysis. By formulating the scheme presentation module for threshold level classification, differentiated presentation schemes can be provided to users according to different similarities. This can better meet the needs of different users and improve user satisfaction. Threshold level classification helps to improve the efficiency of similarity comparison, so that the system can quickly judge and provide users with corresponding presentation schemes. This helps to improve the system's response speed and user experience. As the similarity threshold level increases or decreases, the solution can be flexibly expanded or optimized to provide a more comprehensive and accurate similarity comparison and presentation solution formulation, thereby providing users with differentiated presentation solutions based on the similarity threshold level between standard language data and educational resource videos, thereby improving the system's efficiency and user experience.

[0174] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0175] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A video education resource query method based on natural language processing, characterized in that: The steps include: S1: Receiving search data statistics: The user inputs the search term into the search platform, collects the search platform and the search term data, and stores the search platform and the search term data independently; S2: Retrieval data analysis: Perform natural language processing on the user's retrieval data, where natural language processing includes semantic information analysis and key information analysis of the retrieval terms, classify the retrieval terms after natural language processing into attribute categories, and mark them as standard language data; S3: Retrieval platform matching: Confirm the video resource library in the retrieval platform, analyze the videos in the video resource library by language category, and classify the videos in the video resource library by category according to the analysis results; S4: Video resource presentation: importing the standard language data into the video resource library for category matching, confirming the similarity of the standard language data matching, importing the similarity threshold into the presentation rule to formulate a presentation plan, and presenting the formulated plan to the display terminal; In S2, the standard language data is imported into the video resource library for category matching, and the similarity of the standard language data matching is confirmed, which also includes: Match the attribute categories of the standard language data with the attribute categories of the educational resource videos in the video resource library; Perform curve data conversion on standard language data and educational resource videos with the same attribute category; Overlapping the curve data of the standard language data with the curve data of the educational resource video; The similarity between the standard language data and the educational resource video is determined based on the range of the overlapping area when the curves overlap; Among them, the more overlapping areas, the higher the similarity between the standard language data and the educational resource video; Overlapping the curve data of the standard language data with the curve data of the educational resource video to obtain an overlapping area; Based on the curve characteristics of the standard language data and the curve characteristics of the educational resource video in the overlapping area, the area S of the overlapping area is calculated according to the following formula; Where n represents the number of intersections between the curve data of the standard language data and the curve data of the educational resource video, S 0i Indicates S ai and S bi The smaller of ai represents the area corresponding to the curve feature of the standard language data corresponding to the i-th intersection, S bi represents the area corresponding to the curve feature of the educational resource video corresponding to the i-th intersection, and ΔSr represents the reference standard area difference; Based on the area of ​​the overlapping region, combined with the curve data of the standard language data, the similarity P between the standard language data and the educational resource video is calculated according to the following formula; Among them, S V It represents the similarity reference area corresponding to the category attributes of the standard language data and the educational resource video. e represents a natural constant with a value of 2.

72. σ The matching weight of the curve data representing the standard language data and the curve data of the educational resource video and the semantic information is (0, 1).

2. The video education resource query method based on natural language processing according to claim 1 is characterized in that: The search platform and search term data in S1 are collected, including: The user first logs in to the search platform and enters the search data in the search engine of the search platform; Confirm the IP address of the search platform where the user logs in; Import the retrieval platform IP address into the address database for security verification; Among them, security verification is to verify the IP address of the retrieval platform with the IP in the address database; When the IP address of the retrieval platform is verified to be qualified, the retrieval platform is a safe platform; When the IP address verification of the search platform fails, the search platform is considered an abnormal platform and will not accept search terms. The security platform and the search terms in the security platform search engine are stored separately.

3. The video education resource query method based on natural language processing according to claim 2 is characterized in that: Perform natural language processing on the user's search data in S2, including: Before retrieving data for natural language processing, the retrieved data is preprocessed; Among them, the search data is first cleaned, and the data cleaning is to eliminate abnormal values, missing values ​​and duplicate values ​​in the search data; The search data after data cleaning is converted, wherein the data conversion includes conversion of numerical data and text data in the search data, conversion of categorical variables and continuous variables, and normalization of features; Normalizing the search data after data conversion, wherein data normalization is to adjust the dimension and range of the search data to be consistent; After data normalization, the retrieval data with data preprocessing completed is obtained.

4. The video education resource query method based on natural language processing according to claim 3 is characterized in that: Semantic information analysis and key information analysis of search terms in S2 also include: The retrieval data after data preprocessing is analyzed for semantic information and key information in turn; Among them, the search data is first subjected to semantic information analysis, and the semantic information analysis is to extract feature vectors from the search data through the bag-of-words model; The extracted feature vectors are used for model training through a computer learning model; Perform inference on the retrieval data after model training, and obtain the semantic information of the retrieval data after inference; In the inference process, the feature vector of the retrieved data is predicted by the model, and the semantic information of the retrieved data is obtained after the prediction. The semantic information includes sentiment analysis, topic classification, and question-answering system; Finally, the semantic information of the retrieved data is obtained; Then, the search data is subjected to key information extraction. The key information extraction uses the bag-of-words model to extract keyword-related features in the search data, where the keyword-related features are text, images, and audio. Import the extracted keyword information into the semantic information for keyword selection; The selected keywords are used as the key information for retrieving data.

5. The video education resource query method based on natural language processing according to claim 4 is characterized in that: The retrieval data after model training is inferred, and semantic information of the retrieval data is obtained after inference, including: The retrieval data after model training is processed in a data planning manner according to the target reasoning language conversion strategy to obtain the target retrieval data; Based on the semantic analysis dimension, determining a marking rule for the target retrieval data, marking the target retrieval data according to the marking rule, and obtaining target marked data; Based on the marking features of the target marked data, feature vectors are extracted from the target marked data to obtain a feature vector group corresponding to each semantic analysis dimension, and a dimension matrix is ​​constructed based on the feature vector group; Based on the matrix characteristics of the dimension matrix, determine the reasoning scheme for different semantic analysis dimensions, and process the initial dimension analysis reasoning model based on the reasoning scheme to obtain the target dimension analysis reasoning model; Input the dimension matrix into the corresponding target dimension analysis and reasoning model, and output the initial reasoning information and the associated data and supplementary data related to the dimension matrix; Obtaining initial semantic information of the retrieved data based on the initial reasoning information; Determining whether the correlation between the initial semantic information and the associated data and the supplementary data is greater than a preset correlation; If so, determining that the initial semantic information is the final semantic information of the retrieved data; Otherwise, based on the difference between the correlation between the initial semantic information and the associated data and supplementary data and the preset correlation, the adjustment weight is determined, and based on the adjustment weight, combined with the reference semantic information of the associated data and supplementary data, the initial semantic information is adjusted to obtain the final semantic information of the retrieved data.

6. The video education resource query method based on natural language processing according to claim 4 is characterized in that: The search terms that have been processed by natural language in S2 are classified into attribute categories, including: Determine the attribute category of the search data based on the key information, where the attribute category is the subject of the search data, including science, art, history, politics, geography, and language; Import the key information into the subject database for attribute category confirmation, and obtain the standard language data after the attribute category confirmation of the retrieved data is completed.

7. The video education resource query method based on natural language processing according to claim 6 is characterized in that: Perform language category analysis on the videos in the video resource library in S3, and classify the videos in the video resource library according to the analysis results, including: Confirm each educational resource video in the video resource library; Then, the language and text of each educational resource video are obtained; Extract keywords from the language and text of educational resource videos; Performing attribute category judgment on the keywords extracted from the language and text of the educational resource video, wherein the attribute category of the keywords extracted from the language and text in the educational resource video is consistent with the attribute category of the search term; Determine the final category of the educational resource video according to the keyword attribute category extracted from the language and text in the educational resource video; And the educational resource videos of the same category are uniformly classified.

8. A video education resource query system based on natural language processing, applied in the video education resource query method based on natural language processing as claimed in claim 6, comprising: Develop a program presentation module to: Classify the similarity between the standard language data and the educational resource videos by threshold level; Among them, the threshold levels are divided into primary level, secondary level and tertiary level; The presentation plan is formulated according to the threshold level classification of similarity; The formulated plans are presented on the display terminal in sequence according to the threshold levels.

Images