A standard intelligent recommendation method based on text mining

Abstract

The present invention relates to the technical field of text data analysis, and in particular to a standard intelligent recommendation method based on text mining. The present invention divides an online teaching video into a number of playing time periods according to the playing time progress, determines whether the online teaching video is a feature action explicit video according to the fluctuation degree of the number of feature actions in the number of playing time periods, and determines the feature action explicit time period, determines whether the playing time corresponding to the question asked by the user in the online teaching video is within the feature action explicit time period, so as to determine the index category of the question, obtains a number of question phrases by performing word segmentation on the question text, and determines the response mode of content recommendation according to the index category of the question, thereby realizing the analysis of the difficulty of the question content according to the timing of the user's question, and quickly recommends the corresponding teaching content according to the difference in the difficulty of the question, thereby avoiding the information overload of the online teaching recommendation content and improving the personalization of the online teaching recommendation content.

Description

Technical Field

[0001] The present invention relates to the technical field of text data analysis, and in particular to a standard intelligent recommendation method based on text mining. Background Art

[0002] With the rapid development of Internet technology, online teaching has become an important means to promote balanced education. Online teaching needs to carry out educational activities adaptively according to the learning conditions of different students. The online teaching method that integrates text mining aims to improve the accuracy and efficiency of online teaching in recommending teaching content based on questions from different students through deep learning and natural language processing technology. Scenario-based, content-based and detailed analysis of users' questions is the key to achieving personalized recommendations for users in online teaching at this stage.

[0003] Chinese patent publication number: CN106354844A, discloses a service combination package recommendation system and method based on text mining, including step 1: collecting historical service combination data to construct a data set, and constructing the historical service combination and the functional description of the historical service combination in the data set into a service database; step 2: dividing the functional description of the historical service combination in the service database into a number of sentence blocks according to rhetorical relations; step 3: constructing a sentence relationship tree for each service combination to obtain the relationship between different sentence blocks; step 4: obtaining the relationship between different services; step 5: statistically analyzing the relationship between services to generate a relationship matrix; step 6: decomposing user needs into sentence blocks, and recommending the top K services for each sentence block, where K is a positive integer; step 7: semantically matching the services in each sentence block in combination with the relationship matrix to obtain a service package recommended to the user.

[0004] However, the prior art still has the following problems:

[0005] The existing technology does not take into account that during the user's learning process of online teaching content, due to differences in user learning abilities, the content and timing of users' questions are different. The existing technology cannot analyze the difficulty of the question content according to the timing of the user's question, and cannot quickly recommend corresponding teaching content based on the difference in question difficulty, resulting in a lack of personalization in the recommended teaching content. Summary of the invention

[0006] To this end, the present invention provides a standard intelligent recommendation method based on text mining to overcome the problems that the prior art cannot analyze the difficulty of question content according to the timing of user questions, and cannot quickly recommend corresponding teaching content according to the difference in question difficulty.

[0007] To achieve the above object, the present invention provides a standard intelligent recommendation method based on text mining, which comprises:

[0008] Step S1, dividing the online teaching video into several playing periods according to the playing time progress, and counting the number of characteristic actions in each playing period in the historical records, wherein the number of characteristic actions is determined based on the number of pauses and replays in the playing period;

[0009] Step S2, determining whether the online teaching video is a characteristic action explicit video based on the fluctuation degree of the characteristic action frequency in several playback time periods, and determining the characteristic action explicit time period of the characteristic action explicit video based on the characteristic action frequency in each playback time period;

[0010] Step S3, determining whether the playback time corresponding to the question raised by the user in the online teaching video is within the characteristic action explicit time period, so as to determine the index category of the question;

[0011] Step S4, obtaining a question text, and segmenting the question text to obtain a number of question phrases, determining a response method for content recommendation based on the index category of the question, including determining the corresponding playback time when each question phrase text coincides with the subtitle text of the online teaching video, and recommending the playback time corresponding to each question phrase to the user;

[0012] Or, recommending relevant teaching content to the user based on each question phrase, and determining the content type of the relevant teaching content based on the understanding difficulty representation coefficient;

[0013] The comprehension difficulty representation coefficient is calculated based on the number of characteristic actions of a single user in the characteristic action explicit period in the historical records, and the content type includes a text content type and a video content type.

[0014] Furthermore, the step S1 further includes determining the number of pauses and the number of replays within the playback period, and determining the sum of the number of pauses and the number of replays as the number of characteristic actions.

[0015] Furthermore, in step S2, the standard deviation of the number of characteristic actions is calculated according to formula (1):

[0016]

[0017] In formula (1), Nc is the standard deviation of the number of characteristic actions, N i is the number of characteristic actions in the i-th playback period, i = 1, 2, 3…n, n is the number of playback periods divided by the online teaching video, N av It is the average number of characteristic actions in several playback periods.

[0018] Furthermore, in step S2, the process of determining whether the online teaching video is a feature action explicit video is as follows:

[0019] Comparing the standard deviation of the number of characteristic actions with a preset standard deviation threshold of the number of characteristic actions;

[0020] If the standard deviation of the number of characteristic actions is greater than the standard deviation threshold of the number of characteristic actions, the online teaching video is determined to be a characteristic action explicit video.

[0021] Furthermore, in step S2, the process of determining the characteristic action explicit time period of the characteristic action explicit video includes:

[0022] Comparing the number of characteristic actions in the playback period with a reference value of the number of characteristic actions;

[0023] If the characteristic action frequency is greater than the characteristic action frequency reference value, determining the playback period as the characteristic action explicit period;

[0024] The reference value of the number of characteristic actions is determined based on the average value of the number of characteristic actions in several playback time periods.

[0025] Furthermore, in step S3, the process of determining the index category of the question is:

[0026] If the playback time corresponding to the question raised by the user in the online teaching video is not within the characteristic action explicit time period, then the question is determined to be an internal index category;

[0027] If the playback time corresponding to the question raised by the user in the online teaching video is within the characteristic action explicit time period, the question is determined to be an external index category.

[0028] Furthermore, in step S4, the process of determining the response mode of content recommendation is as follows:

[0029] If the question is an internal index category, determining the corresponding playback time when the text of each question phrase overlaps with the subtitle text of the online teaching video, and recommending the playback time corresponding to each question phrase to the user;

[0030] If the question is of an external index category, relevant teaching content is recommended to the user based on each question phrase, and the content type of the relevant teaching content is determined based on the understanding difficulty representation coefficient.

[0031] Furthermore, in step S4, the step of determining the corresponding playing time when each question phrase text coincides with the subtitle text of the online teaching video includes:

[0032] Step S41, comparing the text similarity of each question phrase text with the subtitle text of the online teaching video;

[0033] Step S42, determining the texts whose text similarity comparison results meet the preset requirements as overlapping texts;

[0034] Step S43, determining the time when the overlapping text appears in the subtitles of the online teaching video as the playback time.

[0035] Furthermore, in step S4, the understanding difficulty representation coefficient is calculated according to formula (2):

[0036] E=λ×n a / n a0 +η×n b / n b0 (2)

[0037] In formula (2), E is the coefficient representing the difficulty of understanding, n a is the maximum number of pauses of a single user during the explicit period of the characteristic action, n a0 is the average number of pauses during several playback periods in the online teaching video, n b is the maximum number of times a single user replays the feature action during the explicit period, n b0 is the average number of replay times in several playback periods of the online teaching video, λ is the weight factor of the number of pauses, and η is the weight factor of the number of replays.

[0038] Furthermore, in step S4, the process of determining the content type of the relevant teaching content includes:

[0039] Comparing the comprehension difficulty representation coefficient with a preset comprehension difficulty representation coefficient threshold;

[0040] If the comprehension difficulty representation coefficient is less than or equal to the comprehension difficulty representation coefficient threshold, determining that the content type of the relevant teaching content is a text content type;

[0041] If the comprehension difficulty representation coefficient is greater than the comprehension difficulty representation coefficient threshold, it is determined that the content type of the relevant teaching content is a video content type.

[0042] Compared with the prior art, the present invention divides the online teaching video into several playback time periods according to the playback time progress, determines whether the online teaching video is a feature-action explicit video by the degree of fluctuation of the number of feature actions in several playback time periods, and determines the feature-action explicit time period, determines whether the playback time corresponding to the question asked by the user in the online teaching video is within the feature-action explicit time period to determine the index category of the question, obtains several question phrases by segmenting the question text, and determines the response method of content recommendation by the index category of the question, thereby realizing the analysis of the difficulty of the question content according to the timing of the user's question, and quickly recommends the corresponding teaching content according to the difference in question difficulty, thereby avoiding information overload of the recommended online teaching content and improving the personalization of the recommended online teaching content.

[0043] In particular, the present invention determines whether an online teaching video is a feature-action explicit video by calculating the degree of fluctuation of the number of feature actions within several playback time periods in the online teaching video. Those skilled in the art can understand that in the process of users watching teaching videos, when encountering knowledge points that are difficult to master, users may pause the teaching video for a long time before and after specific knowledge points, and may repeatedly watch a certain part of the content in the teaching video. These are all manifestations of users trying to understand the knowledge points in the teaching video. The present invention characterizes the distribution of difficult-to-master knowledge points in online teaching videos by calculating the degree of fluctuation of the sum of the number of pauses and the number of replays within several playback time periods in the online teaching video. The greater the degree of fluctuation of the number of pauses and the number of replays, the more concentrated the distribution of difficult-to-master knowledge points in the online education video is in certain playback time periods. In addition, the distribution of difficult-to-master knowledge points in online teaching videos is distinguished.

[0044] In particular, the present invention determines the characteristic action explicit time period of the characteristic action explicit video by the number of characteristic actions in each playback time period. The more frequently the user pauses and replays in certain playback time periods of the online teaching video, the more likely it is that the users have difficulty in mastering the teaching content in these playback time periods. The present invention determines the playback time periods with significantly more characteristic actions in the playback time periods as characteristic action explicit time periods, thereby realizing the screening of the playback time periods with difficult-to-master knowledge points.

[0045] In particular, the present invention determines the index category of the question by determining whether the playback time corresponding to the question asked by the user in the online teaching video is within the characteristic action explicit time period. Those skilled in the art can understand that if the playback time corresponding to the question asked by the user in the online teaching video is not within the characteristic action explicit time period, it means that the user's question is not caused by the knowledge point being difficult to master, but may be caused by the user being distracted and not mastering the content previously explained in the teaching video. At this time, internal indexing is first performed according to the content of the online teaching video to quickly find the video explanation content corresponding to the question asked, and the corresponding time is recommended to the user, so that the user can choose the teaching content that he or she has not listened to carefully before according to the recommendation; if the playback time corresponding to the question asked by the user in the online teaching video is within the characteristic action explicit time period, it means that the user's question is caused by the knowledge point being difficult to master. At this time, internal indexing according to the content of the online teaching video cannot solve the problem that the relevant knowledge points are not fully explained, and it is necessary to timely recommend relevant teaching content from the external index to supplement it, thereby realizing the rapid recommendation of corresponding teaching content according to the difference in question difficulty.

[0046] In particular, the present invention determines the content type of relevant teaching content by calculating the coefficient representing the difficulty of understanding. It can be understood by those skilled in the art that, during a playback period in which knowledge points are difficult to master, the greater the maximum number of video pauses and replays performed by a single user, the greater the difficulty of the knowledge points represented in this playback period. When the difficulty of the knowledge points in the playback period does not exceed the set difficulty, the user's questions can be quickly solved by supplementing the teaching content of the relevant text content type. When the difficulty of the knowledge points in the playback period exceeds the set difficulty, the user's questions can be more comprehensively and intuitively solved by supplementing the teaching content of the relevant video content type. Compared with the teaching content of the text content type, the teaching content of the video content type has a more comprehensive and detailed content explanation, which can help users better master the knowledge points. The teaching content of the text content type can facilitate users to quickly filter out content that helps users understand the knowledge points, thereby avoiding information overload of the recommended online teaching content and improving the personalization of the recommended online teaching content. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 A step diagram of a standard intelligent recommendation method based on text mining according to an embodiment of the present invention;

[0048] Figure 2 A logic flow chart of determining whether an online teaching video is a feature action explicit video according to an embodiment of the present invention;

[0049] Figure 3 A logic flow chart for determining the index category of a question in an embodiment of the present invention;

[0050] Figure 4 FIG. 4 is a diagram showing the steps of determining a playback time according to an embodiment of the present invention. DETAILED DESCRIPTION

[0051] In order to make the objects and advantages of the present invention more clearly understood, the present invention is further described below in conjunction with embodiments; it should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0052] The preferred embodiments of the present invention are described below with reference to the accompanying drawings. It can be understood by those skilled in the art that these embodiments are only used to explain the technical principles of the present invention and are not intended to limit the protection scope of the present invention.

[0053] It should be noted that, in the description of the present invention, terms such as "inside" and "outside" indicating directions or positional relationships are based on the directions or positional relationships shown in the drawings. This is only for the convenience of description and does not indicate or imply that the device or element must have a specific orientation, be constructed and operated in a specific orientation. Therefore, it cannot be understood as a limitation on the present invention.

[0054] See also Figure 1 As shown, it is a step diagram of a standard intelligent recommendation method based on text mining according to an embodiment of the present invention. The standard intelligent recommendation method based on text mining according to the present invention includes:

[0055] Step S1, dividing the online teaching video into several playing periods according to the playing time progress, and counting the number of characteristic actions in each playing period in the historical records, wherein the number of characteristic actions is determined based on the number of pauses and replays in the playing period;

[0056] Step S2, determining whether the online teaching video is a characteristic action explicit video based on the fluctuation degree of the characteristic action frequency in several playback time periods, and determining the characteristic action explicit time period of the characteristic action explicit video based on the characteristic action frequency in each playback time period;

[0057] Step S3, determining whether the playback time corresponding to the question raised by the user in the online teaching video is within the characteristic action explicit time period, so as to determine the index category of the question;

[0058] Step S4, obtaining a question text, and segmenting the question text to obtain a number of question phrases, determining a response method for content recommendation based on the index category of the question, including determining the corresponding playback time when each question phrase text coincides with the subtitle text of the online teaching video, and recommending the playback time corresponding to each question phrase to the user;

[0059] Or, recommending relevant teaching content to the user based on each question phrase, and determining the content type of the relevant teaching content based on the understanding difficulty representation coefficient;

[0060] The comprehension difficulty representation coefficient is calculated based on the number of characteristic actions of a single user in the characteristic action explicit period in the historical records, and the content type includes a text content type and a video content type.

[0061] Specifically, in an embodiment of the present invention, the method of dividing the online teaching video into several playback time periods according to the playback time progress can be to divide the playback time progress according to a preset time interval, and the time interval t is determined according to the total time T of the online teaching video, t=α×T, wherein α is the time interval value factor, and the value range of α can be [0.03, 0.08].

[0062] Specifically, the present invention does not limit the specific method of obtaining the number of pauses and replays of the online teaching video in each playback period. In an embodiment of the present invention, an event listener can be used in the teaching video playback background to capture the number of pauses and replays of the online teaching video by the user. Different programming languages ​​can be selected to monitor the progress bar stop event and move event of the player, and count the number of times the progress bar stop and move events are triggered. In this way, the number of times the progress bar is stopped and replayed within a certain period of time can be known. For example, JavaScript can be used to determine the monitoring processing of Video video playback, pause, progress bar movement and replay events. This technology is widely used in related fields of obtaining video content and viewing behavior, and will not be repeated here.

[0063] Specifically, the present invention does not limit the way of segmenting the question text to obtain several question phrases. Preferably, word segmentation is a basic step in text mining of the question text. Preferably, for the word segmentation of the Chinese question text, a corpus can be used to establish a statistical probability, and the optimal word segmentation can be achieved by calculating the joint distribution probability corresponding to various word segmentation methods. The word segmentation method of the N-gram model can also be selected. In an embodiment of the present invention, the n ltk word segmentation tool can be selected to perform word segmentation on the English text, and the jieba word segmentation tool can be selected to perform word segmentation on the Chinese text. The n ltk word segmentation tool and the jieba word segmentation tool are widely used in existing translation, simultaneous interpretation and speech recognition tools, and will not be repeated here.

[0064] Specifically, in an embodiment of the present invention, a method for determining whether each question phrase text overlaps with the subtitle text of an online teaching video can be to calculate the similarity between each question phrase text and the subtitle text of the online teaching video. Text similarity calculation is an important issue in natural language processing. Preferably, a cosine similarity calculation method can be selected from common text similarity calculation methods. The question phrase text and the subtitle text are segmented, and each segmented word is converted into a vector representation. Finally, the cosine angle between the two vectors is calculated. The smaller the cosine angle, the more similar the two texts are. Two texts whose cosine angles meet preset requirements are determined to be overlapping texts.

[0065] Specifically, in the embodiment of the present invention, the way of recommending the playback time corresponding to each question phrase to the user can be through a pop-up window, which is a prior art and will not be described in detail here.

[0066] Specifically, the present invention does not limit the method of recommending relevant teaching content to users based on each question phrase. The content-based recommendation algorithm can recommend content by analyzing the content description information of the question phrase. Word vectors can be used to represent each phrase, and the similarity between them and existing resource data can be calculated for content recommendation. This technology is widely used in academic information retrieval systems and will not be repeated here.

[0067] Specifically, the step S1 further includes determining the number of pauses and the number of replays within the playback period, and determining the sum of the number of pauses and the number of replays as the number of characteristic actions.

[0068] Specifically, in step S2, the standard deviation of the number of characteristic actions is calculated according to formula (1):

[0069]

[0070] In formula (1), Nc is the standard deviation of the number of characteristic actions, N i is the number of characteristic actions in the i-th playback period, i = 1, 2, 3…n, n is the number of playback periods divided by the online teaching video, N av It is the average number of characteristic actions in several playback periods.

[0071] Specifically, see Figure 2 As shown, it is a logic flow chart of determining whether an online teaching video is a feature action explicit video according to an embodiment of the present invention. In step S2, the process of determining whether the online teaching video is a feature action explicit video is as follows:

[0072] Compare the characteristic action number standard deviation Nc with the preset characteristic action number standard deviation threshold Nc0;

[0073] If the characteristic action number standard deviation Nc is less than or equal to the characteristic action number standard deviation threshold Nc0, it is determined that the online teaching video is not a characteristic action explicit video;

[0074] If the standard deviation Nc of the number of characteristic actions is greater than the standard deviation threshold Nc0 of the number of characteristic actions, the online teaching video is determined to be a characteristic action explicit video.

[0075] In the embodiment of the present invention, preferably, the value of the standard deviation threshold Nc0 of the number of characteristic actions is based on the average value N of the number of characteristic actions in several playback periods. av Determine, its value range is [0.2N av , 0.3N av ].

[0076] Specifically, the present invention determines whether an online teaching video is a feature-action explicit video by calculating the degree of fluctuation of the number of feature actions within several playback time periods in the online teaching video. Those skilled in the art can understand that in the process of users watching teaching videos, when encountering knowledge points that are difficult to master, users may pause the teaching video for a long time before and after specific knowledge points, and may repeatedly watch a certain part of the content in the teaching video. These are all manifestations of users trying to understand the knowledge points in the teaching video. The present invention characterizes the distribution of difficult-to-master knowledge points in online teaching videos by calculating the degree of fluctuation of the sum of the number of pauses and the number of replays within several playback time periods in the online teaching video. The greater the degree of fluctuation of the number of pauses and the number of replays, the more concentrated the distribution of difficult-to-master knowledge points in the online education video is in certain playback time periods. In addition, the distribution of difficult-to-master knowledge points in online teaching videos is distinguished.

[0077] Specifically, in step S2, the process of determining the characteristic action explicit time period of the characteristic action explicit video includes:

[0078] Compare the number of characteristic actions N in the playback period with a reference value of the number of characteristic actions N';

[0079] If the characteristic action number N is less than or equal to the characteristic action number reference value N', it is determined that the playback period is not a characteristic action explicit period;

[0080] If the characteristic action number N is greater than the characteristic action number reference value N', determining the playback period as the characteristic action explicit period;

[0081] The characteristic action number reference value N' is based on the average value N of the characteristic action number in several playback periods. avDetermine that the reference value N' of the characteristic action times is in the range of [1.2N av , 1.25N av ].

[0082] Specifically, the present invention determines the characteristic action explicit time period of the characteristic action explicit video by the number of characteristic actions in each playback time period. The more frequently the user pauses and replays in certain playback time periods of the online teaching video, the more likely it is that the users have difficulty in mastering the teaching content in these playback time periods. The present invention determines the playback time periods with significantly more characteristic actions in the playback time periods as characteristic action explicit time periods, thereby realizing the screening of the playback time periods with difficult-to-master knowledge points.

[0083] Specifically, see Figure 3 As shown, it is a logic flow chart of determining the index category of the question according to an embodiment of the present invention. In the step S3, the process of determining the index category of the question is as follows:

[0084] If the playback time corresponding to the question raised by the user in the online teaching video is not within the characteristic action explicit time period, then the question is determined to be an internal index category;

[0085] If the playback time corresponding to the question raised by the user in the online teaching video is within the characteristic action explicit time period, the question is determined to be an external index category.

[0086] Specifically, the present invention determines the index category of the question by determining whether the playback time corresponding to the question asked by the user in the online teaching video is within the characteristic action explicit time period. Those skilled in the art can understand that if the playback time corresponding to the question asked by the user in the online teaching video is not within the characteristic action explicit time period, it means that the user's question is not caused by the knowledge point being difficult to master, but may be caused by the user being distracted and not mastering the content previously explained in the teaching video. At this time, internal indexing is first performed according to the content of the online teaching video to quickly find the video explanation content corresponding to the question asked, and the corresponding time is recommended to the user, so that the user can choose the teaching content that he or she has not listened to carefully before according to the recommendation; if the playback time corresponding to the question asked by the user in the online teaching video is within the characteristic action explicit time period, it means that the user's question is caused by the knowledge point being difficult to master. At this time, internal indexing according to the content of the online teaching video cannot solve the problem that the relevant knowledge points are not fully explained, and it is necessary to timely recommend relevant teaching content from the external index to supplement it, thereby realizing the rapid recommendation of corresponding teaching content according to the difference in question difficulty.

[0087] Specifically, in step S4, the process of determining the response method of content recommendation is as follows:

[0088] If the question is an internal index category, determining the corresponding playback time when the text of each question phrase overlaps with the subtitle text of the online teaching video, and recommending the playback time corresponding to each question phrase to the user;

[0089] If the question is of an external index category, relevant teaching content is recommended to the user based on each question phrase, and the content type of the relevant teaching content is determined based on the understanding difficulty representation coefficient.

[0090] Specifically, see Figure 4 As shown, it is a step diagram of determining the playback time in an embodiment of the present invention. In step S4, the step of determining the playback time corresponding to when each question phrase text coincides with the subtitle text of the online teaching video includes:

[0091] Step S41, comparing the text similarity of each question phrase text with the subtitle text of the online teaching video;

[0092] Step S42, determining the texts whose text similarity comparison results meet the preset requirements as overlapping texts;

[0093] Step S43, determining the time when the overlapping text appears in the subtitles of the online teaching video as the playback time.

[0094] In an embodiment of the present invention, the question phrase text and the subtitle text can be converted into vector form, and by calculating the vector dot product, the vector length and the cosine similarity, a cosine similarity calculation result of 1 indicates that the question phrase text and the subtitle text are exactly the same, that is, the two texts overlap.

[0095] Specifically, in step S4, the understanding difficulty representation coefficient is calculated according to formula (2):

[0096] E=λ×n a / n a0 +η×n b / n b0 (2)

[0097] In formula (2), E is the coefficient representing the difficulty of understanding, n a is the maximum number of pauses of a single user during the explicit period of the characteristic action, n a0 is the average number of pauses during several playback periods in the online teaching video, n b is the maximum number of times a single user replays the feature action during the explicit period, n b0 is the average number of replay times in several playback periods of the online teaching video, λ is the weight factor of the pause times, η is the weight factor of the replay times, and λ+η=1.

[0098] In an embodiment of the present invention, the values ​​of the pause number weight factor λ and the replay number weight factor η are determined based on the degree of influence of the pause number and the replay number in the historical data calculation on the calculation result. Preferably, the pause number weight factor λ can be set to 0.4, and the replay number weight factor η can be set to 0.6.

[0099] Specifically, in step S4, the process of determining the content type of the relevant teaching content includes:

[0100] Comparing the comprehension difficulty characterization coefficient E with a preset comprehension difficulty characterization coefficient threshold value E0;

[0101] If the comprehension difficulty representation coefficient E is less than or equal to the comprehension difficulty representation coefficient threshold E0, determining that the content type of the relevant teaching content is a text content type;

[0102] If the comprehension difficulty representation coefficient E is greater than the comprehension difficulty representation coefficient threshold E0, it is determined that the content type of the relevant teaching content is a video content type.

[0103] In the embodiment of the present invention, preferably, the value range of the comprehension difficulty characterization coefficient threshold E0 is [1.1, 1.15].

[0104] Specifically, the present invention determines the content type of relevant teaching content by calculating the coefficient representing the difficulty of understanding. It can be understood by those skilled in the art that, during a playback period in which knowledge points are difficult to master, the greater the maximum number of video pauses and replays performed by a single user, the greater the difficulty of the knowledge points represented in this playback period. When the difficulty of the knowledge points in the playback period does not exceed the set difficulty, the user's questions can be quickly solved by supplementing the teaching content of the relevant text content type. When the difficulty of the knowledge points in the playback period exceeds the set difficulty, the user's questions can be more comprehensively and intuitively solved by supplementing the teaching content of the relevant video content type. Compared with the teaching content of the text content type, the teaching content of the video content type has a more comprehensive and detailed content explanation, which can help users better master the knowledge points. The teaching content of the text content type can facilitate users to quickly filter out content that helps users understand the knowledge points, thereby avoiding information overload of the recommended online teaching content and improving the personalization of the recommended online teaching content.

[0105] So far, the technical solutions of the present invention have been described in conjunction with the preferred embodiments shown in the accompanying drawings. However, it is easy for those skilled in the art to understand that the protection scope of the present invention is obviously not limited to these specific embodiments. Without departing from the principle of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will fall within the protection scope of the present invention.

Claims

1. A standard intelligent recommendation method based on text mining, characterized in that: include: Step S1, dividing the online teaching video into several playing periods according to the playing time progress, and counting the number of characteristic actions in each playing period in the historical records, wherein the number of characteristic actions is determined based on the number of pauses and replays in the playing period; Step S2, determining whether the online teaching video is a characteristic action explicit video based on the fluctuation degree of the characteristic action frequency in several playback time periods, and determining the characteristic action explicit time period of the characteristic action explicit video based on the characteristic action frequency in each playback time period; The standard deviation of the characteristic action times is calculated according to formula (1): In formula (1), Nc is the standard deviation of the number of characteristic actions, N i is the number of characteristic actions in the i-th playback period, i=1,2,3…n, n is the number of playback periods divided by the online teaching video, N av is the average number of characteristic actions in several playback periods; Comparing the standard deviation of the number of characteristic actions with a preset standard deviation threshold of the number of characteristic actions; If the standard deviation of the number of characteristic actions is greater than the standard deviation threshold of the number of characteristic actions, the online teaching video is determined to be a characteristic action explicit video; The process of determining the characteristic action explicit time period of the characteristic action explicit video includes: Comparing the number of characteristic actions in the playback period with a reference value of the number of characteristic actions; If the characteristic action frequency is greater than the characteristic action frequency reference value, determining the playback period as the characteristic action explicit period; Wherein, the reference value of the number of characteristic actions is determined based on the average value of the number of characteristic actions in several playback time periods; Step S3, determining whether the playback time corresponding to the question raised by the user in the online teaching video is within the characteristic action explicit time period, so as to determine the index category of the question; The process of determining the index category of the question is: If the playback time corresponding to the question raised by the user in the online teaching video is not within the characteristic action explicit time period, then the question is determined to be an internal index category; If the playback time corresponding to the question raised by the user in the online teaching video is within the characteristic action explicit time period, then the question is determined to be an external index category; Step S4, obtaining a question text, and segmenting the question text to obtain a number of question phrases, determining a response method for content recommendation based on the index category of the question, including determining the corresponding playback time when each question phrase text coincides with the subtitle text of the online teaching video, and recommending the playback time corresponding to each question phrase to the user; Or, recommending relevant teaching content to the user based on each question phrase, and determining the content type of the relevant teaching content based on the understanding difficulty representation coefficient; The comprehension difficulty representation coefficient is calculated based on the number of characteristic actions of a single user in the characteristic action explicit period in the historical records, and the content type includes a text content type and a video content type.

2. The standard intelligent recommendation method based on text mining according to claim 1, characterized in that: The step S1 further includes determining the number of pauses and the number of replays within the playback period, and determining the sum of the number of pauses and the number of replays as the number of characteristic actions.

3. The standard intelligent recommendation method based on text mining according to claim 1, characterized in that: In step S4, the process of determining the response mode of content recommendation is as follows: If the question is an internal index category, determining the corresponding playback time when the text of each question phrase overlaps with the subtitle text of the online teaching video, and recommending the playback time corresponding to each question phrase to the user; If the question is of an external index category, relevant teaching content is recommended to the user based on each question phrase, and the content type of the relevant teaching content is determined based on the understanding difficulty representation coefficient.

4. The standard intelligent recommendation method based on text mining according to claim 3, characterized in that: In step S4, the step of determining the corresponding playback time when each question phrase text coincides with the subtitle text of the online teaching video includes: Step S41, comparing the text similarity of each question phrase text with the subtitle text of the online teaching video; Step S42, determining the texts whose text similarity comparison results meet the preset requirements as overlapping texts; Step S43, determining the time when the overlapping text appears in the subtitles of the online teaching video as the playback time.

5. The standard intelligent recommendation method based on text mining according to claim 3, characterized in that: In step S4, the understanding difficulty representation coefficient is calculated according to formula (2): E=λ×n a / n a0 +η×n b / n b0 (2) In formula (2), E is the coefficient representing the difficulty of understanding, n a is the maximum number of pauses of a single user during the explicit period of the characteristic action, n a0 is the average number of pauses during several playback periods in the online teaching video, n b is the maximum number of times a single user replays the feature action during the explicit period, n b0 is the average number of replay times in several playback periods of the online teaching video, λ is the weight factor of the number of pauses, and η is the weight factor of the number of replays.

6. The standard intelligent recommendation method based on text mining according to claim 5, characterized in that: In step S4, the process of determining the content type of the relevant teaching content includes: Comparing the comprehension difficulty representation coefficient with a preset comprehension difficulty representation coefficient threshold; If the comprehension difficulty representation coefficient is less than or equal to the comprehension difficulty representation coefficient threshold, determining that the content type of the relevant teaching content is a text content type; If the comprehension difficulty representation coefficient is greater than the comprehension difficulty representation coefficient threshold, it is determined that the content type of the relevant teaching content is a video content type.

Images