Online courseware resource management method and system based on artificial intelligence

Abstract

The invention relates to the technical field of online courseware resource management, in particular to an online courseware resource management method and system based on artificial intelligence. Firstly, the quality of the courseware is visually evaluated according to interaction data of the courseware, the total emotion score of the courseware is calculated through the interaction data, for some users who do not publish the interaction data, access data of the users corresponding to the interaction data are removed from all the access data, and therefore preprocessing data are obtained; according to the method, the total recognition degree of the courseware is further calculated through the characteristics of preprocessed data, and finally the total recognition degree and the total emotion score are normalized and fused into the value degree through a specific weight, so that the value degree is compared with a value degree threshold value to recognize the low-quality courseware, subsequent further judgment and processing by workers are facilitated, and the workload of manual screening is reduced; the technical problem that in an existing manual auditing mode, limited manual auditing personnel are difficult to meet the requirement for later low-quality courseware screening is solved.

Description

Technical Field

[0001] This invention relates to the field of online courseware resource management technology, and in particular to an artificial intelligence-based method and system for online courseware resource management. Background Technology

[0002] As online education platforms accumulate courseware resources, the quality of these resources varies greatly. According to Hick's Law, too many inefficient options may lead to user fatigue or even user churn. Therefore, eliminating low-quality courseware from online education platforms can help the platform's brand development, facilitate efficient learning for users, and encourage video creators to provide high-quality courseware.

[0003] Currently, the review of courseware on educational platforms is mainly done through manual review. After users upload courseware, it undergoes initial review by the platform's internal human or AI staff before being uploaded. Subsequently, based on the browsing activity of the courseware, the platform's human reviewers regularly remove low-quality courseware. However, with the dramatic increase in the number of courseware uploads and the limited number of human reviewers, it is difficult to meet the needs of screening low-quality courseware in the later stages. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides an online courseware resource management method and system based on artificial intelligence, which solves the technical problem that the limited number of human reviewers makes it difficult to meet the needs of screening low-quality courseware in the later stages of existing manual review methods.

[0005] To address the aforementioned technical problems, this invention provides the following technical solution: an online courseware resource management method based on artificial intelligence, which specifically includes the following steps: S1. Obtain access and interaction data of online courseware within a certain period of time; S2. Extract adjectives related to the courseware from the interaction data to construct standard binary groups and classify them into several emotion categories, including negative emotions and positive emotions; S3. Calculate the emotional intensity of each adjective in each standard tuple based on the interaction data, and calculate the average emotional intensity of the adjectives in all standard tuples to obtain the total emotional score and value range of the courseware. S4. Remove the access data corresponding to the users of the interaction data from the access data to obtain preprocessed access data, and calculate the total approval score and the range of the total approval score of the courseware. S5. Normalize the range and values ​​of the total emotion score and total approval score, and then calculate the value of the courseware based on the normalized total emotion score and total approval score and the corresponding weights. S6. Determine whether the value score is lower than the value score threshold; If so, output a message indicating that the courseware is of abnormal value, and then end; If not, then the process ends.

[0006] Preferably, step S2 specifically includes the following steps: S21. Obtain a negative emotion word set containing negative emotion words and a positive emotion word set containing positive emotion words; S22. Set the corresponding prompt words in the non-comment data to negative and positive emotions, and set their dependency relationship with the courseware settings; S23. Sequentially obtain the nouns and adjectives in each sentence of each comment data under the courseware to obtain the word segmentation phrases of each comment data, and obtain the word segmentation dependency relationship of each sentence; S24. Based on the syntactic dependency relations of each comment data, construct binary groups from nouns and adjectives that have dependency relations in the word segmentation phrases and are located in the same sentence; S25. Select nouns to refer to the binary pairs of this courseware to obtain standard binary pairs; S26. Based on the word set to which the adjectives in each standard tuple in the interaction data belong, determine the emotional category of each adjective in relation to the courseware. If the adjective in a standard binary tuple belongs to the set of negative emotion words, then the adjective belongs to a negative emotion. If the adjective in a standard binary tuple belongs to the set of positive emotion words, then the adjective belongs to the positive emotion.

[0007] Preferably, step S23 specifically includes the following steps: S231. Sequentially obtain the syntactic dependency relations of each sentence in each comment data under the courseware, and set the start character and end character according to the first character, last character and punctuation mark of the comment data; S232. Extract the text between each adjacent start character and end character and the text between adjacent end characters to obtain each sentence, and input the initial sequence composed of each sentence into the word segmentation extraction model to obtain the word segmentation group of each comment data; The expression for the initial sequence is: ; The expression for word segmentation phrases is: ; In the above formula, This represents any comment data. This represents the i-th sentence in the comment data, which contains n sentences in total. This represents the word segmentation of the comment data. This represents the output result after inputting the i-th sentence of the initial sequence into the word segmentation and extraction model. ; S233. Sequentially determine whether there are only adjectives among the several word segments corresponding to the first sentence in the word segmentation phrase; If so, add the comment object of the comment data to the first sentence of the segmented word group and set the dependency relationship; If not, proceed to step S24.

[0008] Preferably, step S3 specifically includes the following steps: S31. Obtain the number of times each word of each adjective in each standard tuple in the interaction data appears in the negative emotion word set and the positive emotion word set, respectively. S32. Calculate the degree of negative emotion and the degree of positive emotion for each adjective based on the frequency of each word in the negative emotion word set and the positive emotion word set. S33. Subtract the negative emotional intensity from the positive emotional intensity of the adjective to obtain the emotional intensity of the adjective; S34. Determine whether the emotional intensity of adjectives classified as positive emotions is greater than 0, and whether the emotional intensity of adjectives classified as negative emotions is less than 0. If so, then remove the adjective from the standard tuple; If not, then retain the adjective within the standard tuple; S35. Calculate the average of the emotional intensity of all adjectives within all standard tuples to obtain the total emotional score of the courseware, and obtain the range of values ​​for the total emotional score.

[0009] Preferably, step S4 specifically includes the following steps: S41. Remove the access data corresponding to the user of the interaction data from the access data to obtain the first access data. Then remove the first access data whose courseware browsing time is less than the browsing time threshold to obtain the preprocessed access data. S42. Based on the courseware viewing time and effective time of the first viewing in the preprocessed access data, calculate the first approval rating of each user for the courseware; the formula for calculating the first approval rating is: ; In the above formula, This indicates the initial level of approval for the courseware. This indicates the duration of the first viewing of the courseware. Indicates the effective duration of the courseware; S43. Calculate each user's second level of approval for the courseware based on the number of times the user accesses the courseware; the formula for calculating the second level of approval is: ; In the above formula, This indicates the second level of approval. The frequency coefficient, This indicates the number of times a user accesses the courseware; S44. Calculate the average of the second level of approval for the courseware from each customer to obtain the total approval level of the courseware.

[0010] Preferably, step S41 specifically includes the following steps: S411. Remove the access data corresponding to the user of the interaction data from the access data to obtain the first access data; S412. Mark the first access data of the user's first access to the courseware as first access data, and mark the first access data of the user's second access to the courseware after the first viewing as repeated access data. S413. Set a first browsing duration threshold of 2 minutes and a second browsing duration threshold of 30 seconds; S414. Remove first-access data for courseware browsing time that is less than the first browsing time threshold; S415. Remove duplicate access data where the courseware browsing time is less than the second browsing time threshold; S416. Mark the first access data that has not been removed as preprocessed access data.

[0011] Preferably, step S5 specifically includes the following steps: S51. Normalize the range of total emotion score and the range of total approval score; S52. Calculate the value of the courseware based on the normalized total emotion score and total approval rating; the formula for calculating the value is: ; In the above formula, Indicates the value of the courseware. Indicates the sentiment coefficient. This represents the approval rating. This represents the overall emotional score of the courseware. This indicates the overall approval rating of the courseware.

[0012] Preferably, in step S6, the value threshold is set as follows: S61. Set the initial value threshold; S62. Identify abnormal samples containing several courseware items through steps S2-S5 and the initial value threshold. S63. The abnormal samples of courseware are manually classified into courseware to be removed from the shelves and courseware to be kept on the shelves. S64. Obtain the maximum value of the courseware to be removed from the shelves and the minimum value of the courseware to be kept on the shelves; S65. Calculate the average of the maximum value of the courseware to be removed and the minimum value of the courseware to be kept on the shelves, and use it as the value threshold.

[0013] Preferably, in step S32, the formula for calculating the degree of negative emotion of the adjective is: ; The formula for calculating the positive emotional intensity of adjectives is: ; In the above formula, and These respectively indicate the degree of negative and positive emotion conveyed by the adjective. and Let represent the number of times the j-th word of the adjective appears in the negative emotion word set and the positive emotion word set, respectively. The total number of occurrences of this adjective is... Each character.

[0014] The present invention also provides an online courseware resource management system, including a processor and a memory, wherein the memory is used to store a computer program, and the computer program, when executed by the processor, implements the online courseware resource management method.

[0015] By employing the above technical solution, the present invention provides an online courseware resource management method and system based on artificial intelligence, which has at least the following beneficial effects: 1. This invention first evaluates the quality of courseware based on interactive data. It calculates the total sentiment score of the courseware using the interactive data. For users who do not post interactive data, the access data of users corresponding to the interactive data is removed from all access data to obtain preprocessed data. The characteristics of the preprocessed data are used to further calculate the overall approval of the courseware. Finally, the overall approval and the overall sentiment score are normalized and then fused with specific weights to form a value score. This value score is then compared with a value score threshold to identify low-quality courseware, facilitating further manual judgment and processing.

[0016] 2. This invention recognizes that when users post comments, they may be commenting on courseware or on other people's evaluations. Therefore, by identifying the interaction data related to the courseware, standard binary pairs can be extracted, which facilitates the subsequent calculation of the total sentiment score.

[0017] 3. This invention uses a set of positive emotion words to quantify the emotional intensity of adjectives, and uses the range of positive emotion intensity and the range of negative emotion intensity to filter adjectives with positive and negative emotions, thereby ensuring the accuracy of the standard binary set used to calculate the total emotion score.

[0018] 4. In order to determine the overall acceptance of courseware based on traffic statistics, this invention first removes the access data of users corresponding to the interaction data to avoid duplicate calculations. Secondly, by setting corresponding browsing duration thresholds for single and repeated accesses by individual users, the effective duration is filtered and used as data for subsequent calculation of the overall acceptance, which helps to improve the credibility of the overall acceptance.

[0019] 5. This invention sets a value threshold and updates it automatically, thereby adjusting the value threshold according to the results of manual review. This increases the probability of screening out low-quality courseware that meets the manual settings and reduces the labor intensity of manual screening. Attached Figure Description

[0020] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments of this application and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings: Figure 1 This is a flowchart of the online courseware resource management method based on artificial intelligence according to the present invention. Detailed Implementation

[0021] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. This will allow for a full understanding of how the present application uses technical means to solve technical problems and achieve technical effects, and to facilitate its implementation.

[0022] The current manual review method for user-uploaded courseware uses a combination of artificial intelligence and human intervention to obtain the content and identify sensitive information. Courseware without sensitive content and containing educational material is generally approved and uploaded to the platform. After a period of time, based on data such as page views, the platform uses either human or artificial intelligence for further review. However, this method still has many shortcomings. For example, some courseware targets a niche audience but receives overwhelmingly positive reviews from most visitors, potentially leading to AI misidentification as low-quality. Conversely, some courseware may have high page views, but poor video quality causes almost everyone to exit immediately, potentially resulting in misidentification. Manual screening, given the tens of thousands of courseware on the platform, would be extremely time-consuming. To mitigate the shortcomings of existing AI and manual review methods, this paper proposes an AI-based online courseware resource management method, which includes the following steps: Access and interaction data of online courseware directly reflect its quality. Therefore, it is necessary to acquire and process this data to extract elements that objectively reflect the quality of the online courseware. Access data includes the users who accessed the courseware, the number of times the courseware was accessed by each user, the effective duration of the courseware, and the viewing time of each user's access. Interaction data includes comment data and non-comment data. Non-comment data includes characters such as "like," "dislike," "favorite," and "donate" set within the platform. When a user clicks on these characters, a corresponding instruction is triggered. Comment data consists of comments posted by users under the courseware. In addition, this method acquires access and interaction data of online courseware within a certain period of time, rather than access and interaction data for all time since the courseware was uploaded. The purpose is that some courseware is time-sensitive, popular for a certain period of time, but its effectiveness almost disappears after a period of time, so it needs to be cleared. The length of this period of time can be determined according to the server's storage capacity. For example, if the maximum storage capacity is 5 years, then this period of time should be set to be less than 5 years.

[0023] Interaction data includes both comment and non-comment data, thus reflecting users' intuitive feelings about the courseware. Users often use emotionally charged words when expressing their feelings; therefore, adjectives related to the courseware are extracted from the interaction data to construct standard binary groups, which are then categorized into several emotion categories, including negative and positive emotions, reflecting the degree of liking or disliking for the courseware. However, when users post comments, they may be commenting on the courseware itself or on other people's evaluations. Therefore, it is necessary to identify comments on the courseware as data for subsequent calculations, while negative and positive emotions in comments on other people's evaluations do not reflect their emotions towards the courseware. The specific steps include the following: First, in the comment data, there are certain standards for determining whether the words used by users are positive or negative emotions and the degree of emotion. The determination method is as follows: Obtain a negative emotion word set containing negative emotion words and a positive emotion word set containing positive emotion words. The negative emotion word set and the positive emotion word set can be obtained from DUTIR Chinese sentiment lexicon, open-source sentiment lexicons in the field of natural language processing (such as BosonNLP, HowNet, and NTUSD sentiment dictionary).

[0024] For non-comment data, such as common words like "like," "favorite," "donate," and "dislike," which are generally considered positive or negative emotional terms, we directly set the corresponding prompts in the non-comment data to negative and positive emotions, and establish a dependency relationship with the courseware. Typically, "like," "favorite," and "donate" are set to positive relationships, while "dislike" is set to negative emotions.

[0025] After clarifying the criteria for identifying negative and positive emotional words, it is necessary to sequentially obtain the nouns and adjectives in each sentence of each comment data under the courseware to obtain the word segmentation phrases for each comment data, and then obtain the dependency relations of each sentence's word segmentation. In some sentences, nouns may not exist. For example, if someone posts "very good" as a comment data, the object of their comment is the courseware. Therefore, it is necessary to establish a dependency relation with the courseware. Dependency relations are a core concept in syntactic analysis, used to describe the grammatical connections between words in a sentence. By establishing directed relations between words, it reveals the dependencies and grammatical functions between sentence components, and can clearly reveal the grammatical structure and semantic logic of the sentence. It is a natural language processing task, and commonly used algorithms include BERT+BiAffine and Stack-LSTM. However, in the courseware comments, some comments are comments on other comments, and some sentences are extremely simple, which may lead to errors in dependency relations. Therefore, the existing algorithm is supplemented, specifically including the following steps: When obtaining the syntactic dependency relations of each sentence, the first step is to segment the sentences. The following method is provided for sentence segmentation: Sequentially obtain the syntactic dependency relations of each sentence in each comment data entry under the courseware, and set start and end marks based on the first and last characters and punctuation marks of the comment data. According to general language usage habits, set start and end marks at the first and last characters. Additionally, generally set end marks at periods, exclamation marks, and question marks; other punctuation marks are not marked.

[0026] After marking the start and end symbols, the sentence segmentation of each sentence is clear. The next step is to extract words from each sentence. This is done by extracting the text between each adjacent start and end symbol and between each adjacent end symbol to obtain each sentence. The initial sequence composed of each sentence is then input into the word segmentation extraction model to obtain the word groups for each comment data. The expression for the initial sequence is: ; The expression for word segmentation phrases is: ; In the above formula, This represents any comment data. This represents the i-th sentence in the comment data, which contains n sentences in total. This represents the word segmentation of the comment data. This represents the output result after inputting the i-th sentence of the initial sequence into the word segmentation and extraction model. Commonly used word segmentation and extraction models include jieba segmentation and GPT model.

[0027] Based on the habits of online comments, some users may only use words containing adjectives to express their emotional feelings towards the object being commented on. Therefore, it is necessary to distinguish such sentence segments and determine the dependency relationship. Thus, it is necessary to determine whether there are only adjectives in the several segments corresponding to the first sentence in the word segmentation group. If so, add the comment object of the comment data to the first sentence of the segmented word group and set the dependency relationship. For example, if someone posts "very good" as comment data and the object of their comment is courseware, then establish a dependency relationship between "very good" and courseware.

[0028] If not, then based on the syntactic dependency relations of each comment data, nouns and adjectives with dependency relations in the segmented word groups that are located in the same sentence are constructed into tuples. However, for different tuples, the entity sometimes does not refer to the courseware. Therefore, it is necessary to extract the tuples whose entity is courseware, that is, to select the tuples whose nouns refer to the courseware. For example, if someone comments "the course is very good", then the noun "course" refers to the entity "courseware". The method used is the coreference resolution task in natural language processing, which can be implemented by recurrent neural networks, convolutional neural networks or Transformer models, etc., to obtain standard tuples. The form of a standard tuple can be <courseware, very good>.

[0029] After obtaining each standard tuple, we can classify the emotional categories of the adjectives in each standard tuple in the interactive data according to the negative emotional vocabulary and the classification criteria mentioned above. That is, we can determine the emotional category of each adjective for the courseware based on the negative emotional vocabulary set and the positive emotional vocabulary set. If the adjective in a standard binary tuple belongs to the set of negative emotion words, then the adjective belongs to a negative emotion. If the adjective in a standard binary tuple belongs to the set of positive emotion words, then the adjective belongs to the positive emotion.

[0030] After determining whether the adjectives in each standard tuple of the interaction data belong to the negative or positive emotion category, it is necessary to determine their emotional intensity. For example, regarding "like" and "like very much," the latter clearly indicates a deeper emotion. The emotional intensity of each adjective within each standard tuple is calculated based on the interaction data, and the average emotional intensity of all adjectives within all standard tuples is calculated to obtain the total emotional score and value range of the courseware. For instance, if two people post comments such as "good" and "very good," it can be clearly determined that the latter is more likely to appreciate the courseware. Therefore, it is necessary to differentiate between them and assess their emotional intensity, and calculate the overall evaluation level of all users for the courseware, i.e., the total emotional score. This application uses the following method to calculate its emotional intensity and total emotional score: First, obtain the frequency of each word of each adjective in each standard tuple of the interaction data within the negative sentiment word set and the positive sentiment word set, respectively. Then, based on the frequency of each word of the adjective in the negative emotion word set and the positive emotion word set, the negative emotion degree and positive emotion degree of each adjective are calculated respectively; The formula for calculating the degree of negative emotion of an adjective is: ; The formula for calculating the positive emotional intensity of adjectives is: ; In the above formula, and These respectively indicate the degree of negative and positive emotion conveyed by the adjective. and Let represent the number of times the j-th word of the adjective appears in the negative emotion word set and the positive emotion word set, respectively. The total number of occurrences of this adjective is... Each character; When an adjective contains both positive and negative emotional words, the degree of emotion between the two may differ. For example, for the word "bittersweet", the degree of emotion of the adjective is obtained by subtracting the degree of negative emotion from the degree of positive emotion. During the calculation process, there may be a very small probability of anomalies. For example, if an adjective is defined as a positive emotion word but the calculated emotion level is negative, it indicates that the emotion level calculation is inaccurate and the previous emotion level calculation method cannot be fully applied. Therefore, it is necessary to identify such adjectives. The specific method is as follows: determine whether the emotion level of adjectives classified as positive emotions is greater than 0, and whether the emotion level of adjectives classified as negative emotions is less than 0. If the emotion level of a positive emotion adjective is less than 0, or the emotion level of an adjective classified as negative emotions is greater than 0, it indicates a calculation anomaly. Since the calculation of the total emotion score for a courseware is not determined by individual adjectives but rather by evaluating the emotion levels of all adjectives related to the courseware, this method is used to address the fact that a small number of adjectives are not suitable for this calculation method. These adjectives are removed to reduce the computational load. Alternatively, such words could be assigned a separate emotion level value. If the adjective is determined to be abnormal, then the adjective in the standard tuple is removed; otherwise, the adjective in the standard tuple is retained.

[0031] Next, based on the calculated emotional level, the average emotional level of all adjectives within all standard pairs is calculated to obtain the total emotional score of the courseware, and the range of the total emotional score is obtained. The calculation of the total emotional score is explained below. If there are 5 standard pairs that all contain adjectives, and there are a total of 7 adjectives, then there are 2 standard pairs that contain 2 adjectives, or 1 standard pair that contains 3 adjectives. In this case, the emotional level of all adjectives is added together to obtain a sum, and then this sum is divided by the number of standard pairs, 5, to obtain the total emotional score.

[0032] Of all users who viewed the courseware, only a portion expressed their feelings about it. This was already calculated using interaction data in the previous section. However, the emotional feedback obtained through interaction data is more accurate and intuitive. Therefore, to improve the accuracy of the calculation, in subsequent calculations using access data, the access data of users already calculated will be removed. The overall approval rating of users who viewed the courseware but did not express their feelings will be calculated to reduce errors and avoid duplicate calculations. By removing access data corresponding to users with interaction data, preprocessed access data is obtained, and the overall approval rating and its range are calculated. The method for calculating the overall approval rating using access data is explained in detail below: Access data corresponding to users with interactive data was removed from the access data to obtain the first access data. Then, the first access data with a browsing time less than a threshold was removed to obtain preprocessed access data. During the preprocessing process, it was found that some users repeatedly browsed the same content. Analysis of customer browsing activity revealed that customers were more targeted in subsequent browsing, primarily searching for content in specific locations within the courseware. This indicates that the courseware held significant value for the user, and therefore, the browsing time per visit cannot be simply positively correlated with subsequent recognition calculations. Furthermore, the study found that some users might have accidentally clicked on links to the courseware. The approximate browsing time when users encountered uninteresting courseware was statistically analyzed. Users would quickly exit during the first visit, and during subsequent visits, or exit after quickly searching and finding no relevant content. This type of data is invalid and does not improve user recognition of the courseware. Two browsing time thresholds were set to remove this portion of the access data. The specific method is as follows: First, access data corresponding to users with interactive data is removed from the access data to obtain the first access data. Then, the first access data of a user's first access to the courseware is marked as first access data, and the first access data of a user's second access to the courseware after the first viewing is marked as duplicate access data. According to the viewing habits of network users, a first viewing time threshold of 2 minutes and a second viewing time threshold of 30 seconds can be set. Then, based on the set first and second viewing time thresholds, first access data with a viewing time of less than the first viewing time threshold is removed, as this data may be due to users accidentally clicking on the first viewing time threshold. Duplicate access data with a viewing time of less than the second viewing time threshold is also removed, as this data may be due to users entering the courseware while searching for information but finding that it is not the information they need. Finally, the first access data that has not been removed is marked as preprocessed access data, thus completing the data filtering for subsequent calculations.

[0033] The longer a user watches the courseware and the more times they watch it, the more relevant the courseware is to them, indicating higher quality. Therefore, based on the initial viewing time and effective duration of the courseware in the preprocessed access data, we first calculate each user's initial approval rating of the courseware. The formula for calculating the initial approval rating is: ; In the above formula, This indicates the initial level of approval for the courseware. This indicates the duration of the first viewing of the courseware. Indicates the effective duration of the courseware; Next, the second level of approval for each user's courseware is calculated based on the number of times the user accesses the courseware and the first level of approval for each access; the formula for calculating the second level of approval is: ; In the above formula, This indicates the second level of approval. The frequency coefficient, This indicates the number of times a user accesses the courseware. By setting the arctangent function, the maximum value of the second approval score for a single user can be set, which can prevent some users from maliciously inflating the approval score of the courseware.

[0034] After obtaining the second approval rating from a single user, it is necessary to calculate the second approval rating of all users for the courseware to ensure the objectivity of the data. Here, the total approval rating of the courseware is calculated by averaging the second approval ratings of each user.

[0035] Because the calculated total emotion score and total approval score have inconsistent dimensions, the range and values ​​of the total emotion score and total approval score are normalized to ensure consistency of dimensions. Then, the value of the courseware is calculated based on the normalized total emotion score and total approval score and their corresponding weights. The specific method is as follows: First, the ranges of the total emotion score and the total approval score are normalized to ensure consistency of measurement. Then, the value of the courseware is calculated based on the normalized total emotion score and total approval score. The calculation formula is as follows: ; In the above formula, Indicates the value of the courseware. Indicates the sentiment coefficient. This represents the approval rating. This represents the overall emotional score of the courseware. This indicates the overall approval rating of the courseware.

[0036] In the final stage, a value threshold needs to be set. Based on whether the calculated value of the courseware is lower than the threshold, a final judgment is made regarding whether the courseware quality is abnormal. Specifically, if the value of the courseware is lower than the threshold, a message indicating abnormal value is output; otherwise, the judgment of the next courseware can proceed. However, to further ensure the rationality of the value threshold and enable it to adaptively adjust according to the requirements of manual review, while also preventing the initial value threshold from being unreasonable, this process does not require manual adjustment and is more convenient. The method for setting the value threshold is as follows: First, the operations and maintenance personnel need to set an initial value threshold. Then, they need to calculate the value of the courseware through the steps mentioned above and identify several abnormal samples with a value lower than the initial value threshold. After identifying the abnormal samples, the courseware in the abnormal samples is manually classified into courseware to be removed and courseware to be kept on the shelves. Courseware to be removed is courseware with low quality, and courseware to be kept on the shelves is courseware with quality that meets the requirements. Then, the maximum value of the courseware to be removed and the minimum value of the courseware to be kept on the shelves are obtained. At this time, the value threshold should theoretically be located between the maximum value of the courseware to be removed and the minimum value of the courseware to be kept on the shelves. Therefore, the average of the two is calculated and used as the value threshold.

[0037] This invention first evaluates the quality of courseware based on interactive data. It calculates the overall sentiment score of the courseware using this data. For users who do not post interactive data, preprocessed data is obtained by removing the access data of users corresponding to their interactive data from all access data. The overall approval rating of the courseware is then calculated based on the characteristics of this preprocessed data. Finally, the overall approval rating and overall sentiment score are normalized and then fused with specific weights to form a value score. This value score is then compared with a value threshold to identify low-quality courseware, facilitating further manual judgment and processing. For example, it can determine whether a courseware is low-quality, allowing for direct removal or notification of the uploading user for modification or removal.

[0038] The present invention also provides an online courseware resource management system, including a processor and a memory, wherein the memory is used to store computer programs, and when the computer programs are executed by the processor, they implement an artificial intelligence-based online courseware resource management method.

[0039] Those skilled in the art will understand that all or part of the steps in the methods of the above embodiments can be implemented by a program instructing related hardware. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Moreover, this application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0040] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. Since the above embodiments are substantially similar to the method embodiments, their descriptions are relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0041] The above embodiments provide a detailed description of the present invention. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A method for managing online courseware resources based on artificial intelligence, characterized in that, This management method specifically includes the following steps: S1. Obtain access data and interaction data of online courseware within a certain period of time, wherein the interaction data includes comment data and non-comment data; S2. Extract adjectives related to the courseware from the interaction data to construct standard binary groups and classify them into several emotion categories, including negative emotions and positive emotions; S3. Calculate the emotional intensity of each adjective in each standard tuple based on the interaction data, and calculate the average emotional intensity of the adjectives in all standard tuples to obtain the total emotional score and value range of the courseware. S4. Remove the access data corresponding to the users of the interaction data from the access data to obtain preprocessed access data, and calculate the total approval score and the range of the total approval score of the courseware. S5. Normalize the range and values ​​of the total emotion score and total approval score, and then calculate the value of the courseware based on the normalized total emotion score and total approval score and the corresponding weights. S6. Determine whether the value score is lower than the value score threshold; If so, output a message indicating that the courseware is of abnormal value, and then end; If not, then the process ends.

2. The online courseware resource management method according to claim 1, characterized in that, Step S2 specifically includes the following steps: S21. Obtain a negative emotion word set containing negative emotion words and a positive emotion word set containing positive emotion words; S22. Set the corresponding prompt words in the non-comment data to negative and positive emotions, and set their dependency relationship with the courseware settings; S23. Sequentially obtain the nouns and adjectives in each sentence of each comment data under the courseware to obtain the word segmentation phrases of each comment data, and obtain the word segmentation dependency relationship of each sentence; S24. Based on the syntactic dependency relations of each comment data, construct binary groups from nouns and adjectives that have dependency relations in the word segmentation phrases and are located in the same sentence; S25. Select nouns to refer to the binary pairs of this courseware to obtain standard binary pairs; S26. Based on the word set to which the adjectives in each standard tuple in the interaction data belong, determine the emotional category of each adjective in relation to the courseware. If the adjective in a standard binary tuple belongs to the set of negative emotion words, then the adjective belongs to a negative emotion. If the adjective in a standard binary tuple belongs to the set of positive emotion words, then the adjective belongs to the positive emotion.

3. The online courseware resource management method according to claim 2, characterized in that, Step S23 specifically includes the following steps: S231. Sequentially obtain the syntactic dependency relations of each sentence in each comment data under the courseware, and set the start character and end character according to the first character, last character and punctuation mark of the comment data; S232. Extract the text between each adjacent start character and end character and the text between adjacent end characters to obtain each sentence, and input the initial sequence composed of each sentence into the word segmentation extraction model to obtain the word segmentation group of each comment data; The expression for the initial sequence is: ; The expression for word segmentation phrases is: ； In the above formula, This represents any comment data. This represents the i-th sentence in the comment data, which contains n sentences in total. This represents the word segmentation of the comment data. This represents the output result after inputting the i-th sentence of the initial sequence into the word segmentation and extraction model. ; S233. Sequentially determine whether there are only adjectives among the several word segments corresponding to the first sentence in the word segmentation phrase; If so, add the comment object of the comment data to the first sentence of the segmented word group and set the dependency relationship; If not, proceed to step S24.

4. The online courseware resource management method according to claim 1, characterized in that, Step S3 specifically includes the following steps: S31. Obtain the number of times each word of each adjective in each standard tuple in the interaction data appears in the negative emotion word set and the positive emotion word set, respectively. S32. Calculate the degree of negative emotion and the degree of positive emotion for each adjective based on the frequency of each word in the negative emotion word set and the positive emotion word set. S33. Subtract the negative emotional intensity from the positive emotional intensity of the adjective to obtain the emotional intensity of the adjective; S34. Determine whether the emotional intensity of adjectives classified as positive emotions is greater than 0, and whether the emotional intensity of adjectives classified as negative emotions is less than 0. If so, then remove the adjective from the standard tuple; If not, then retain the adjective within the standard tuple; S35. Calculate the average of the emotional intensity of all adjectives within all standard tuples to obtain the total emotional score of the courseware, and obtain the range of values ​​for the total emotional score.

5. The online courseware resource management method according to claim 1, characterized in that, Step S4 specifically includes the following steps: S41. Remove the access data corresponding to the user of the interaction data from the access data to obtain the first access data. Then remove the first access data whose courseware browsing time is less than the browsing time threshold to obtain the preprocessed access data. S42. Based on the courseware viewing time and effective time of the first viewing in the preprocessed access data, calculate the first approval rating of each user for the courseware; the formula for calculating the first approval rating is: ； In the above formula, This indicates the initial level of approval for the courseware. This indicates the duration of the first viewing of the courseware. Indicates the effective duration of the courseware; S43. Calculate each user's second level of approval for the courseware based on the number of times the user accesses the courseware; the formula for calculating the second level of approval is: ； In the above formula, This indicates the second level of approval. The frequency coefficient, This indicates the number of times a user accesses the courseware; S44. Calculate the average of the second level of approval for the courseware from each customer to obtain the total approval level of the courseware.

6. The online courseware resource management method according to claim 5, characterized in that, Step S41 specifically includes the following steps: S411. Remove the access data corresponding to the user of the interaction data from the access data to obtain the first access data; S412. Mark the first access data of the user's first access to the courseware as first access data, and mark the first access data of the user's second access to the courseware after the first viewing as repeated access data. S413. Set a first browsing duration threshold of 2 minutes and a second browsing duration threshold of 30 seconds; S414. Remove first-access data for courseware browsing time that is less than the first browsing time threshold; S415. Remove duplicate access data where the courseware browsing time is less than the second browsing time threshold; S416. Mark the first access data that has not been removed as preprocessed access data.

7. The online courseware resource management method according to claim 1, characterized in that, Step S5 specifically includes the following steps: S51. Normalize the range of total emotion score and the range of total approval score; S52. Calculate the value of the courseware based on the normalized total emotion score and total approval rating; the formula for calculating the value is: ； In the above formula, Indicates the value of the courseware. Indicates the sentiment coefficient. This represents the approval rating. This represents the overall emotional score of the courseware. This indicates the overall approval rating of the courseware.

8. The online courseware resource management method according to claim 1, characterized in that, In step S6, the value threshold is set as follows: S61. Set the initial value threshold; S62. Identify abnormal samples containing several courseware items through steps S2-S5 and the initial value threshold. S63. The abnormal samples of courseware are manually classified into courseware to be removed from the shelves and courseware to be kept on the shelves. S64. Obtain the maximum value of the courseware to be removed from the shelves and the minimum value of the courseware to be kept on the shelves; S65. Calculate the average of the maximum value of the courseware to be removed and the minimum value of the courseware to be kept on the shelves, and use it as the value threshold.

9. The online courseware resource management method according to claim 4, characterized in that, In step S32, the formula for calculating the degree of negative emotion of an adjective is: ； The formula for calculating the positive emotional intensity of adjectives is: ； In the above formula, and These respectively indicate the degree of negative and positive emotion conveyed by the adjective. and Let represent the number of times the j-th word of the adjective appears in the negative emotion word set and the positive emotion word set, respectively. The total number of occurrences of this adjective is... Each character.

10. A system for implementing the online courseware resource management method according to any one of claims 1-9, characterized in that, It includes a processor and a memory, the memory being used to store a computer program, which, when executed by the processor, implements the online courseware resource management method as described in any one of claims 1-9.

Images