Short video addiction control method tolerance test system based on age group division

Abstract

The invention relates to the field of short video anti-addiction control, in particular to a short video addiction control method tolerance test system based on age group division. The system comprises an agent unit used for acquiring a video request message and recording resources based on a video response message; the request interception unit is used for obtaining a user identifier and a video resource identifier requested by the user according to the video request message, and generating a first vector based on association of video request history of the user; the response unit is used for obtaining a user response message according to the request message of the user and obtaining a first associated recommended video based on the user request; the response rewriting unit is used for determining a second associated recommended video based on the first vector and determining a third associated recommended video in combination with the first associated recommended video; and the tolerance determination unit is used for determining the tolerance of the user to the video recommendation according to the historical record of the video request message of the user. Therefore, the problem of testing the tolerance of people of different age levels is solved.

Description

Technical Field

[0001] This invention relates to the field of short video addiction prevention and control, and in particular to a tolerance testing system for short video addiction control methods based on age group segmentation. Background Technology

[0002] With the development of information technology, short video platforms have rapidly gained popularity due to their rich and diverse content and convenient user experience, becoming one of the main sources of daily entertainment. However, short videos are not so friendly to people of different ages who lack self-control, especially teenagers and retirees. Unhealthy lifestyle habits or excessively commercialized advertisements seriously threaten their physical and mental well-being and their wallets.

[0003] However, given the current situation, completely banning these groups from watching short videos is unrealistic, and simply recommending videos they haven't followed or aren't interested in would only backfire. To address this, this invention proposes a tolerance testing system for short video addiction control based on age group segmentation. By classifying the tolerance levels of different age groups, targeted short video recommendations are made, subtly reducing the viewing time, habits, and content viewed by those lacking self-control, thereby protecting users' physical and mental health and promoting the healthy development of the industry. Summary of the Invention

[0004] In order to address the problem of testing tolerance in different age groups, at least one aspect and advantage of the present invention will be set forth in part in the description which follows, or may be apparent from the description, or may be obtained by practicing the subject matter of this disclosure.

[0005] According to a first aspect of the present invention, a tolerance testing system for a short video addiction control method based on age group segmentation includes:

[0006] The proxy unit is used to obtain video request messages and record resources based on video response messages;

[0007] The request interception unit is used to obtain the user identifier and the video resource identifier requested by the user based on the video request message, obtain the user's historical requested videos based on the user identifier, and generate a first vector based on the association of the user's historical requested videos.

[0008] The response unit obtains a user response message based on the user's request message. The user response message includes the resource information requested by the user and a first associated recommended video obtained based on the user's request.

[0009] The response rewriting unit determines the second associated recommended video based on the first vector, and determines the third associated recommended video based on the first and second associated recommended videos, and generates a video response message based on the resource information requested by the user and the third associated recommended video;

[0010] The tolerance determination unit is used to determine the user's tolerance for video recommendations based on the user's historical video request messages.

[0011] According to one embodiment of the present invention, the generation of the first vector includes:

[0012] The user's video request history is obtained based on the user identifier. The user's video request history includes the time when the user requested the video and the vector information corresponding to the requested video. The vector information of the requested video is generated based on the summary or classification information of the requested video.

[0013] A first requested video sequence is constructed based on the distribution of the time of the user's requested video within a first window. Each element of the first requested video sequence includes a weight and vector information corresponding to the requested video distributed within the time window.

[0014] Based on each element of the first requested video sequence, obtain the vector information corresponding to the time window;

[0015] The first vector is obtained based on the vector information corresponding to each time window.

[0016] According to an embodiment of the present invention, the first vector acquisition process further includes:

[0017] The vector information corresponding to each element in the first requested video sequence is merged in chronological order from newest to oldest to obtain the second vector;

[0018] The second vector is dimensionality reduced to match the dimension of the vector corresponding to the video, thus obtaining the third vector;

[0019] The first vector is obtained based on the third vector and the vector corresponding to the video resource identifier requested by the user.

[0020] The weights of each element in the first requested video sequence are consistent.

[0021] According to one embodiment of the present invention, the vector information corresponding to each element in the first requested video sequence is merged in chronological order from newest to oldest to obtain a first vector;

[0022] The weight of each element in the first requested video sequence tends to decrease in chronological order from newest to oldest.

[0023] According to one embodiment of the present invention, the cosine approximation between the vector corresponding to the first associated recommended video and the first vector is higher than a first threshold, and the cosine approximation between the vector corresponding to the user requested resource information is lower than the first threshold.

[0024] According to one embodiment of the present invention, the third associated video is generated in the following manner:

[0025] Cluster the vectors corresponding to the second associated recommended video to obtain cluster centers;

[0026] Merge the first associated recommended video and the first associated recommended video list to obtain the fourth associated video list;

[0027] The fourth associated video list is sorted according to Euclidean distance based on the cluster centers, and the first m items of the sorted fourth associated video list are used as the third video list, where m is the length of the first associated recommended video list.

[0028] According to one embodiment of the present invention, the user's tolerance for video recommendations is obtained based on the changing trend of user request messages compared to the historical records of video request messages.

[0029] According to one embodiment of the present invention, the user's tolerance for video recommendations is obtained in the following manner:

[0030] Iterate through the history of user video request messages to obtain each video request message and several videos requested by the user before that video request message;

[0031] Furthermore, for each video request message, the video vector corresponding to the user's requested resource is obtained based on the identifier of the video resource in the video request message; the video vectors corresponding to the user's previous n videos are normalized to obtain the historical vector;

[0032] Based on the difference between the video vector and the historical vector corresponding to the user's requested resource, a fourth vector is obtained and added to the video difference list;

[0033] An anomaly detection algorithm is used to process the video difference list, identify the video request message corresponding to the anomaly, and determine the duration of continuous playback before the video request message as the video tolerance.

[0034] According to one embodiment of the present invention, the anomaly detection algorithm is a clustering algorithm.

[0035] According to one embodiment of the present invention, a method is used to determine a person's tolerance for video recommendations based on the temporal distribution of a user's video request messages.

[0036] The beneficial effects of this invention are as follows: By configuring a proxy unit, a request interception unit, a response unit, a response rewriting unit, and a tolerance determination unit, it is possible to test the tolerance of people of different age groups to short videos, providing platform operators with targeted optimization suggestions, thereby more effectively preventing and controlling short video addiction and protecting users' physical and mental health. At the same time, this method can also provide a scientific basis for relevant policymakers, promoting the healthy development of the short video industry. Attached Figure Description

[0037] Figure 1 A unit virtual structure diagram of a tolerance test system for short video addiction control methods based on age group segmentation is shown. Detailed Implementation

[0038] The contents of this disclosure will now be discussed with reference to several exemplary embodiments.

[0039] According to one embodiment of the present invention, a tolerance testing system for a short video addiction control method based on age group segmentation includes:

[0040] The proxy unit is used to obtain video request messages and record resources based on video response messages;

[0041] The request interception unit is used to obtain the user identifier and the video resource identifier requested by the user based on the video request message, obtain the user's historical requested videos based on the user identifier, and generate a first vector based on the association of the user's historical requested videos.

[0042] The response unit obtains a user response message based on the user's request message. The user response message includes the resource information requested by the user and a first associated recommended video obtained based on the user's request.

[0043] The response rewriting unit determines the second associated recommended video based on the first vector, and determines the third associated recommended video based on the first and second associated recommended videos, and generates a video response message based on the resource information requested by the user and the third associated recommended video;

[0044] The tolerance determination unit is used to determine the user's tolerance for video recommendations based on the user's historical video request messages.

[0045] In this embodiment, a tolerance testing system for short video addiction control methods based on age group division is proposed. The system includes the following units.

[0046] Agent unit: mainly used to obtain video request messages and video response messages. The user terminal can load videos and display other recommended videos based on the video response messages.

[0047] The request interception unit is mainly used to obtain the user identifier and the identifier of the video resource requested by the user based on the video request message; at the same time, it obtains the user's historical requested videos based on the user identifier and generates a first vector based on the relevant content of the user's historical requested videos.

[0048] The response unit can obtain the user's response message based on the user's video request message. The user's response message contains the resource information requested by the user, as well as the first associated recommended video determined based on the user's request.

[0049] The response rewriting unit is used to determine the second associated recommended video based on the first vector, then merge the first associated recommended video and the second associated recommended video to obtain the third associated recommended video, and generate a video response message based on the resource information requested by the user and the third associated recommended video.

[0050] The tolerance determination unit is mainly used to determine the user's tolerance for recommended videos based on the user's historical video request messages.

[0051] Specifically, the video response message is used by the user terminal to load videos and display other recommended videos.

[0052] Specifically, the first recommended videos are determined based on the user's response message, which is obtained based on the user's request message. Therefore, the first recommended videos are actually a list of videos that the user is interested in or follows.

[0053] Specifically, the second set of recommended videos is obtained based on the first vector, which is derived from the user's video request history determined by the user's identifier. Therefore, the first vector is a set of feature vectors of videos that the user is interested in or follows. By using feature vectors that are different from or even opposite to the feature vectors in the first vector, a list of video types that the user is not interested in or even finds annoying can be obtained; this is the second set of recommended videos.

[0054] Specifically, by merging the first and second associated recommended video lists, a third associated recommended video list is obtained. This third list contains videos that the user is interested in, has not followed, is not interested in, dislikes, or hates. Based on the user's historical video request messages, corresponding videos are selected from this third list for recommendation. The user's tolerance for the recommended videos can be determined by analyzing the time spent on the videos or the distribution of times the user re-requests them.

[0055] Specifically, the tolerance in this application is an objective data point based on the user's choice and viewing of videos. It is independent of the personnel receiving the test and is mainly based on the user's viewing of videos with similar or different features from the first associated recommended videos.

[0056] Specifically, the tolerance testing system for short video addiction control based on age group segmentation in this invention can intermittently recommend videos with different characteristics from those in the first associated video recommendation list, based on a third associated video recommendation list. This accurately assesses the tolerance of users of different age groups to short video addiction control measures, providing platform operators with targeted optimization suggestions, thereby more effectively preventing and controlling short video addiction and protecting users' physical and mental health. Simultaneously, this method can also provide scientific evidence for relevant policymakers, promoting the healthy development of the short video industry. For example, by rewriting user request-response videos based on age group, for children, videos related to finance, history, instrumental music (containing only background images), and news can be recommended; while for adults, cartoons for young children, news with low click rates, and low-resolution short videos can be recommended.

[0057] According to one embodiment of the present invention, the generation of the first vector includes:

[0058] The user's video request history is obtained based on the user identifier. The user's video request history includes the time when the user requested the video and the vector information corresponding to the requested video. The vector information of the requested video is generated based on the summary or classification information of the requested video.

[0059] A first requested video sequence is constructed based on the distribution of the time of the user's requested video within a first window. Each element of the first requested video sequence includes a weight and vector information corresponding to the requested video distributed within the time window.

[0060] Based on each element of the first requested video sequence, obtain the vector information corresponding to the time window;

[0061] The first vector is obtained based on the vector information corresponding to each time window.

[0062] This embodiment describes a process for generating the first vector. First, the user's video request history is obtained based on the user identifier. Based on the user's video request history, the time of the requested video and the vector information corresponding to the requested video can be obtained. Then, based on the distribution of the time of the user's requested video within a first window period, a first requested video sequence is constructed based on the videos within a certain time window. Each element of this sequence includes a weight and the vector information corresponding to the video within that time window. Finally, the vector information corresponding to the time window is obtained based on each element of the first requested video sequence and is used as the first vector.

[0063] Specifically, on short video apps, based on a user's request, the backend server prepares a series of videos related to that request. After the user watches the first video requested or skips the first video, they can continue to see the requested videos. Therefore, generally speaking, within a certain time window after a request is made, the videos the user sees are all related to the request. However, after the user has viewed several videos, the server may push an advertisement-related video or other recommended videos. These videos are unrelated to the user's request. When the user sees these videos, they may choose to skip them or resubmit the request, and may not watch the videos in their entirety.

[0064] Specifically, the time distribution of user-requested videos within the first window is obtained. Based on the time distribution, videos within a subsequent time window after the user's request time are selected as the first requested video sequence. Correspondingly, each element in the first video request sequence represents a video that the user is interested in or follows. The vector information corresponding to the requested video within that time window is the set of feature vectors of the videos the user is interested in or follows, which is the first vector. Therefore, the first vector is the set of feature vectors of the videos the user is interested in or follows.

[0065] Specifically, the vector information of a video is generated based on the summary or classification information of the requested video, including the video's category, subcategory, domain, duration, and whether it's landscape or portrait mode. For example, if user A likes to watch historical videos and is interested in videos related to the Three Kingdoms period, then the first vector of the video for user A might contain features such as the classification information "historical" and the summary information containing keywords related to the Three Kingdoms and its famous generals. Similarly, user B likes to watch anime, and within the anime category, is particularly interested in Chinese fantasy / cultivation anime adapted from novels, typically watching landscape videos around 20 minutes long.

[0066] Specifically, the first vector obtained by the method in this embodiment can effectively express the features corresponding to the videos that the user is interested in or pays attention to, making the bounding range of the first vector more accurate.

[0067] According to an embodiment of the present invention, the first vector acquisition process further includes:

[0068] The vector information corresponding to each element in the first requested video sequence is merged in chronological order from newest to oldest to obtain the second vector;

[0069] The second vector is dimensionality reduced to match the dimension of the vector corresponding to the video, thus obtaining the third vector;

[0070] The first vector is obtained based on the third vector and the vector corresponding to the video resource identifier requested by the user.

[0071] The weights of each element in the first requested video sequence are consistent.

[0072] This embodiment introduces a method to make the first vector more accurate. First, the vector information of each element in the first requested video sequence is merged according to time from newest to oldest to obtain a second vector. Second, the second vector is dimensionality-reduced to match the dimension of the vector corresponding to the video, resulting in a third vector. Finally, based on the third vector and the vector of the video corresponding to the user-requested video resource identifier, the feature vectors that coexist in both are selected to obtain a more accurate first vector. Specifically, by merging the vector information corresponding to each video in the first requested video sequence with consistent weights, a second vector with more comprehensive video feature vectors can be obtained, ensuring that video features of interest to the user are not missed.

[0073] Specifically, the second vector is a merging of the vector information corresponding to all elements within the first requested video sequence. Since the vector information corresponding to each element may have different dimensions, the dimension of the merged second vector is higher than the dimension of the original vector corresponding to the video. Therefore, reducing the dimension of the second vector to the dimension of the vector corresponding to the video yields the third vector. This facilitates subsequent calculations and removes some redundant features.

[0074] Specifically, the video resource identifier requested by the user is obtained based on the video request message. Then, by using its corresponding vector information, the feature types that the user is interested in or follows can be further identified. Simultaneously, by matching this vector with the feature types in the third vector to obtain the first vector, the final first vector becomes more accurate, removing some interfering features introduced by advertisements or recommendations. This, in turn, makes it easier to control whether the second set of recommended videos deviates from the user's preferences.

[0075] According to one embodiment of the present invention, the vector information corresponding to each element in the first requested video sequence is merged in chronological order from newest to oldest to obtain a first vector;

[0076] The weight of each element in the first requested video sequence tends to decrease in chronological order from newest to oldest.

[0077] In this embodiment, another method to make the first vector more accurate is introduced. Specifically, each element in the first requested video sequence is merged from newest to oldest, wherein the weight of each element decreases in chronological order from newest to oldest.

[0078] Specifically, in the actual video viewing scenario, to assess user tolerance, the system typically uses third-party related recommended videos as recommendations. These third-party recommended videos include those that deviate from the user's focus or interests. When users watch videos they are interested in, they tend to watch them for longer than for videos that deviate from the user's focus. Therefore, the closer the video is to the end of the first window, the newer it is, and the more representative it is of the characteristics of videos that the user is interested in. For example, within a 5-minute first window, if the user watches 5 videos and makes 5 requests in the first minute, then in the second minute, the system will adjust the video content based on the user's requests. In the second minute, the user may watch 4 complete videos and make 4 requests due to the adjustment. In the third minute, this might be 3 complete videos and 3 requests, and so on. Newer videos are more likely to match the characteristics of the user's focus or interests.

[0079] Therefore, assigning weights that decrease from newest to oldest videos can more evenly reflect the characteristics of videos that users are interested in or pay attention to. The resulting first vector contains more accurate features, which in turn allows for a more precise assessment of the deviation of the second set of recommended videos.

[0080] According to one embodiment of the present invention, the cosine approximation between the vector corresponding to the first associated recommended video and the first vector is higher than a first threshold, and the cosine approximation between the vector corresponding to the user requested resource information is lower than the first threshold.

[0081] In this embodiment, the first vector and the user-requested resource information are defined by the cosine approximation of the first associated recommended video.

[0082] Specifically, setting a first threshold restricts the cosine similarity between the first vector and the first associated recommended video to be higher than the first threshold. This results in a higher degree of similarity between the two, which is more in line with practical application scenarios. Restricting the cosine similarity between the vector corresponding to the user's requested resource information and the first vector to be lower than the first threshold limits the features of the response video to those with lower similarity to the first vector, thus facilitating a more objective test of the user's tolerance.

[0083] According to one embodiment of the present invention, the third associated video is generated in the following manner:

[0084] Cluster the vectors corresponding to the second associated recommended video to obtain cluster centers;

[0085] Merge the first associated recommended video and the first associated recommended video list to obtain the fourth associated video list;

[0086] The fourth associated video list is sorted according to Euclidean distance based on the cluster centers, and the first m items of the sorted fourth associated video list are used as the third video list, where m is the length of the first associated recommended video list.

[0087] This embodiment describes the specific method for determining the third associated video. First, the vectors corresponding to the second associated recommended videos are clustered, and cluster centers are obtained. Then, the lists of first and second associated recommended videos are merged to obtain a fourth associated video list. Finally, the fourth associated video list is sorted according to the cluster centers to obtain the third associated video.

[0088] Specifically, the fourth list of recommended videos is sorted according to Euclidean distance based on the cluster centers, and the top m items are used as the third list of recommended videos. The cluster centers here are obtained by clustering the vectors of the second list of recommended videos, which are videos that deviate from the user's focus or interests. Therefore, selecting the top m items after sorting—that is, selecting videos closer to the cluster centers—represents videos that are further removed from the user's focus or interests, making it easier for the user to become fatigued while watching them, thus making it easier to achieve tolerance through the method described in this application.

[0089] According to one embodiment of the present invention, the user's tolerance for video recommendations is obtained based on the changing trend of user request messages compared to the historical records of video request messages.

[0090] According to one embodiment of the present invention, the user's tolerance for video recommendations is obtained in the following manner:

[0091] Iterate through the history of user video request messages to obtain each video request message and several videos requested by the user before that video request message;

[0092] Furthermore, for each video request message, the video vector corresponding to the user's requested resource is obtained based on the identifier of the video resource in the video request message; the video vectors corresponding to the user's previous n videos are normalized to obtain the historical vector;

[0093] Based on the difference between the video vector and the historical vector corresponding to the user's requested resource, a fourth vector is obtained and added to the video difference list;

[0094] An anomaly detection algorithm is used to process the video difference list, identify the video request message corresponding to the anomaly, and determine the duration of continuous playback before the video request message as the video tolerance.

[0095] According to one embodiment of the present invention, the anomaly detection algorithm is a clustering algorithm.

[0096] In this embodiment, the user's tolerance for video recommendations is obtained based on the changing trend of the user request message compared to the historical records of video request messages.

[0097] Specifically, first, the historical records of user video request messages are traversed to obtain each video request message and several videos requested by the user before that video request message. Second, the video vector of the corresponding user-requested resource is obtained based on the tags corresponding to the video resources in the video request message. Third, the sum of the video vectors corresponding to the n videos preceding the user's video request message is normalized to obtain the historical vector. Fourth, a fourth vector is obtained by the difference between the video vector of the user-requested resource and the historical vector, and this fourth vector is added to the video difference list. Finally, the video difference list is processed to obtain the tolerance.

[0098] Specifically, normalization ensures that data from different dimensions in a historical vector can be compared on the same scale.

[0099] Specifically, the video vector of the resource requested by the user represents the feature vector of the videos the user is interested in or follows. The history vector, on the other hand, represents the sum of the feature vectors of the videos the user actually watched. The difference between the two yields a fourth vector, which reflects the feature differences between the currently requested video and the videos the user has watched in the past. These differences are added to the video difference list.

[0100] Specifically, a clustering algorithm is used to process the video difference list to find outliers, that is, to find the inflection point where user interests change, and to determine the video request message corresponding to the outlier. The continuous playback time of the video before the video request message is used as the tolerance level.

[0101] Specifically, when a user's requested video changes significantly, it inevitably indicates a shift in the user's interests. By comparing the video vector of the requested resource with historical vectors, anomalies can be identified. Identifying the corresponding video request message through these anomalies reveals the continuous playback duration of videos prior to that point. This reflects how long a user is willing to watch video content they are not interested in or who have changed their interests. Using this viewing duration as a tolerance level objectively demonstrates the user's tolerance for videos that deviate from their preferred or more interesting content.

[0102] According to one embodiment of the present invention, a method is used to determine a person's tolerance for video recommendations based on the temporal distribution of a user's video request messages.

[0103] In this embodiment, the user's tolerance for video recommendations is determined based on the temporal distribution of the user's video request messages.

[0104] Specifically, in real-world applications, users tend to watch videos that are within their watchlist or interests for a longer period than videos that deviate from their watchlist or interests. If a video deviates from the user's watchlist or interests, the user will switch frequently, resulting in video request messages. The time distribution of these video request messages can then reflect the user's tolerance for the recommended videos. For example, a user might be interested in football (specifically, football) and FC Barcelona matches in their viewing history. However, the recommended videos might include Real Madrid, as well as Premier League and Bundesliga matches. The user will switch between watching these leagues or teams. In this case, the viewing time distribution will be such that the user spends a significant amount of time watching FC Barcelona matches, during which no video request messages are generated, and the user watches the entire video. However, when watching Premier League or Bundesliga matches, the user will only see the summary before switching videos, resulting in a corresponding increase in video request messages during this period. The time distribution of these video request messages can thus indicate the user's tolerance for the recommended videos.

[0105] Specifically, tolerance determined by the time distribution of user video request messages is more suitable for tolerance testing within a specific user group, such as teenagers aged 10-16 or retirees aged 60-65.

[0106] Specifically, the types of videos requested by different age groups will vary greatly. However, in this invention, the user's tolerance for the recommended videos is determined based on the time distribution of video request messages, taking into account the history of video requests for each age group and user feedback on the videos.

[0107] Specifically, to avoid test failures, the third-party recommended videos also include videos randomly selected from a pre-set pool of videos, which are determined based on the user's age. This ensures that the user's short video list includes videos that are less frequently viewed by the majority of people in that age group, meaning the user is more likely not to watch these videos, thus altering the temporal distribution of the user's video request messages.

[0108] Specifically, the third type of recommended videos also includes videos randomly selected from a set of preset videos. These preset videos are selected based on statistics of the probability of users of different ages watching videos, and videos with fewer than the average number of views (e.g., they can be set to 1 / 5-1 / 200).

[0109] The beneficial effects of this invention are as follows: The system of this application can effectively assess the acceptance and effectiveness of short video addiction control measures among users of different age groups, accurately evaluate the tolerance of users of different age groups to these measures, and provide platform operators with targeted optimization suggestions. This allows for more effective prevention and control of short video addiction, protecting users' physical and mental health. Simultaneously, this method can also provide a scientific basis for relevant policymakers, promoting the healthy development of the short video industry.

Claims

1. A tolerance testing system for short video addiction control methods based on age group segmentation, characterized in that, include: The proxy unit is used to obtain video request messages and record resources based on video response messages; The request interception unit is used to obtain the user identifier and the video resource identifier requested by the user based on the video request message, obtain the user's video request history based on the user identifier, and generate a first vector based on the association of the user's video request history. The response unit obtains a user response message based on the user's request message. The user response message includes the resource information requested by the user and a first associated recommended video obtained based on the user's request. The response rewriting unit determines the second associated recommended video based on the first vector, and determines the third associated recommended video based on the first and second associated recommended videos, and generates a video response message based on the resource information requested by the user and the third associated recommended video; The tolerance determination unit is used to determine the user's tolerance for video recommendations based on the user's historical video request messages.

2. The tolerance testing system for the short video addiction control method based on age group segmentation as described in claim 1, characterized in that, The generation of the first vector includes: The user's video request history is obtained based on the user identifier. The user's video request history includes the time when the user requested the video and the vector information corresponding to the requested video. The vector information of the requested video is generated based on the summary or classification information of the requested video. A first requested video sequence is constructed based on the distribution of the time of the user's requested video within a first window. Each element of the first requested video sequence includes a weight and vector information corresponding to the requested video distributed within the time window. Based on each element of the first requested video sequence, obtain the vector information corresponding to the time window; The first vector is obtained based on the vector information corresponding to each time window.

3. The tolerance testing system for the short video addiction control method based on age group segmentation as described in claim 2, characterized in that, The first vector acquisition process also includes: The vector information corresponding to each element in the first requested video sequence is merged in chronological order from newest to oldest to obtain the second vector; The second vector is dimensionality reduced to match the dimension of the vector corresponding to the video, thus obtaining the third vector; The first vector is obtained based on the third vector and the vector corresponding to the video resource identifier requested by the user. The weights of each element in the first requested video sequence are consistent.

4. The tolerance testing system for the short video addiction control method based on age group segmentation as described in claim 2, characterized in that, The vector information corresponding to each element in the first requested video sequence is merged in chronological order from newest to oldest to obtain the first vector; The weight of each element in the first requested video sequence tends to decrease in chronological order from newest to oldest.

5. The tolerance testing system for the short video addiction control method based on age group segmentation as described in claim 1, characterized in that, The cosine approximation between the vector corresponding to the first associated recommended video and the first vector is higher than the first threshold, and the cosine approximation between the vector corresponding to the user requested resource information is lower than the first threshold.

6. The tolerance testing system for the short video addiction control method based on age group segmentation as described in claim 1, characterized in that, The third associated recommended video is generated in the following manner: Cluster the vectors corresponding to the second associated recommended video to obtain cluster centers; Merge the first and second related recommended video lists to obtain the fourth related recommended video list; The fourth associated recommended video list is sorted according to Euclidean distance based on the cluster centers. The first m items of the sorted fourth associated recommended video list are used as the third associated recommended video list, where m is the length of the first associated recommended video.

7. The tolerance testing system for the short video addiction control method based on age group segmentation as described in claim 1, characterized in that, User tolerance for video recommendations is determined based on the historical trend of user request messages compared to video request messages.

8. The tolerance testing system for the short video addiction control method based on age group segmentation as described in claim 7, characterized in that, The user's tolerance for video recommendations is obtained in the following way: Iterate through the history of user video request messages to obtain each video request message and several videos requested by the user before that video request message; Furthermore, for each video request message, the video vector corresponding to the user-requested resource is obtained based on the identifier of the video resource within the video request message; Normalize the sum of the video vectors corresponding to the n videos before the user's requested video to obtain the historical vector; Based on the difference between the video vector and the historical vector corresponding to the user's requested resource, a fourth vector is obtained and added to the video difference list; An anomaly detection algorithm is used to process the video difference list, identify the video request message corresponding to the anomaly, and determine the duration of continuous playback before the video request message as the video tolerance.

9. The tolerance testing system for the short video addiction control method based on age group segmentation as described in claim 8, characterized in that, The anomaly detection algorithm is a clustering algorithm.

10. The tolerance testing system for the short video addiction control method based on age group segmentation as described in claim 1, characterized in that, Used to determine a person's tolerance for video recommendations based on the temporal distribution of their video request messages.

Images