Video-based classification label processing method, device, terminal and medium

Abstract

The present invention discloses a method, device, terminal and medium for processing classification label marks based on video, wherein the method includes: collecting and obtaining active viewing behavior data of users who watch a specified video; confirming the user group data whose liking degree for the specified video is greater than a predetermined value, and screening out the first video data watched by the users of the user group within a specified time; based on the first video data, only retaining the second video data whose liking degree for the user is greater than a predetermined value, and obtaining the video type label corresponding to the second video data; based on the video type label, obtaining a preset number of video type labels of user groups with different characteristics in the user group data as the video type label of the specified video. Compared with the prior art, the present invention can avoid the problem of poor video push effect caused by incorrect label upload by users, and can also help video platforms to quickly and accurately analyze, process and classify massive video resources, improve the speed of automatic labeling, greatly save manual review efficiency, and save costs.

Description

Technical Field

[0001] The present invention relates to the field of video processing technology, and in particular to a video-based classification label processing method, device, terminal and medium. Background Art

[0002] With the development of science and technology, especially the rapid development of Internet technology, various videos are spreading faster and faster through various network platforms.

[0003] At present, due to the increasing amount of Internet video data, the time users spend watching short videos and small videos is also growing rapidly. However, existing technologies cannot quickly and accurately analyze, process, and classify massive video resources, which is sometimes inconvenient for users.

[0004] Therefore, the existing technology still needs to be improved and developed. Summary of the invention

[0005] The main purpose of the present invention is to provide a video-based classification label processing method, device, intelligent terminal and computer-readable storage medium. The present invention can classify and label videos in a simple way, thereby improving the speed of automatic video labeling, automatically classifying and labeling, greatly saving manual review efficiency and costs, while achieving accurate user recommendations and providing convenience for users.

[0006] In order to achieve the above object, the first aspect of the present invention provides a video-based classification label processing method, wherein the method comprises:

[0007] A. Collect and obtain active viewing behavior data of users who have active viewing behavior for a specified video;

[0008] B. Based on the active viewing behavior data, confirm the user group data whose liking degree for the specified video is greater than a predetermined value, and filter out the first video data watched by the user group data in the most recent specified time period;

[0009] C. Based on the filtered first video data, the first video data is cleaned to filter out second video data whose user preference is greater than a predetermined value, and a video type label watched by the user group data in a recently specified time period is obtained;

[0010] D. Based on the obtained video type labels of the user group data watched in the most recent specified time period, a predetermined number of type labels are selected from the type labels of the specified video by different user groups as the final video labels of the specified video, and the label weights owned by the corresponding users are assigned according to the user's preference.

[0011] Optionally, the step of collecting and obtaining active viewing behavior data of users who have an active viewing behavior for a specified video includes:

[0012] According to the user's active viewing behavior indicators, collect and obtain the user's active viewing behavior data for the specified video;

[0013] Based on the user's active viewing behavior data for the specified video, a statistical analysis is performed on the user's liking for the specified video.

[0014] Optionally, the step of confirming the user group data whose liking degree for the specified video is greater than a predetermined value based on the active viewing behavior data, and filtering out the first video data watched by the user group data in the most recent specified time period includes:

[0015] Based on the user preference of each user for the specified video confirmed by the active viewing behavior data, confirm the user group data whose preference for the specified video is greater than a predetermined value;

[0016] Based on the user group data, first video data watched by the user group data in a recent specified time period is filtered out.

[0017] Optionally, the step of performing data cleaning on the first video data to filter out second video data whose user preference is greater than a predetermined value based on the filtered first video data, and obtaining the video type label watched by the user group data in the most recent specified time period includes:

[0018] Based on the first video data that the filtered user group data has watched in the most recent specified time period, data cleaning is performed on the first video data that each user in the user group data has watched in the most recent specified time period, video data whose user preference is less than a predetermined value is cleaned out, and the existing tags of the remaining videos are counted;

[0019] The k video tags with the highest number of appearances are used as the video tags most recently watched by the corresponding user;

[0020] And count the video tags recently watched by each user in the user group data to obtain the video type tags watched by the user group data in the most recent specified time period.

[0021] Optionally, the step of selecting a predetermined number of type tags from type tags of the specified video by different user groups as final video tags of the specified video based on the obtained type tags of the video watched by the user group data in the most recent specified time period, and assigning weights to the tags owned by the corresponding users according to the user's preference degree includes:

[0022] Extracting user features of users who are interested in the specified video, and performing unsupervised clustering to obtain different user groups G1, G2, Gn;

[0023] And assign different weights according to the number of different groups, w1, w2, wn;

[0024] Suppose whether a label appears in G1 is is_label1, where it is 1 if it appears and 0 if it does not appear. The final formula for calculating the label weight is: is_label1*w1+is_label2*w2+...+is_labeln*wn;

[0025] The r video tags with the largest number of tag appearances are used as the video tags of the specified video.

[0026] Optionally, the step of selecting a predetermined number of type tags from type tags of the specified video by different user groups as final video tags of the specified video based on the obtained type tags of the video watched by the user group data in the most recent specified time period, and assigning weights to the tags owned by the corresponding users according to the user's preference degree includes:

[0027] Based on the obtained video type labels of the user group data watched in the most recent specified time period, assigning corresponding weights to the users according to the degree of their active behavior towards the specified videos;

[0028] Clustering users according to the partial features to obtain type labels of the specified video for different user groups;

[0029] A specified number of type tags are selected from type tags of the specified video by different user groups as type tags of the specified video.

[0030] Optionally, the step of selecting a predetermined number of type tags from the type tags of the specified video by different user groups as the final video tags of the specified video based on the obtained type tags of the video watched by the user group data in the most recent specified time period, and assigning weights to the tags owned by the corresponding users according to the user's preference degree, comprises:

[0031] The accuracy of video labels is evaluated, and the weights and label cleaning weights are adjusted based on feedback.

[0032] A second aspect of the present invention provides a video-based classification label processing device, wherein the device comprises:

[0033] A viewing behavior data acquisition module is used to collect and acquire active viewing behavior data of users who have active viewing behavior for a specified video;

[0034] A screening module, for identifying user group data whose liking for the specified video is greater than a predetermined value based on the active viewing behavior data, and screening out first video data watched by the user group data in a recent specified time period;

[0035] A video type label initial confirmation module is used to perform data cleaning on the first video data to filter out second video data whose user preference is greater than a predetermined value based on the first video data filtered out, and obtain the video type label watched by the user group data in the most recent specified time period;

[0036] The final video tag confirmation module is used to select a predetermined number of type tags from the type tags of the specified video by different user groups based on the video type tags watched by the user group data in the most recent specified time period as the final video tags of the specified video, and assign weights to the tags owned by the corresponding users according to the user's preference.

[0037] A third aspect of the present invention provides an intelligent terminal, which includes a memory, a processor, and a video-based classification label marking processing program stored in the memory and executable on the processor, wherein the video-based classification label marking processing program implements any step of the video-based classification label marking processing method when executed by the processor.

[0038] A fourth aspect of the present invention provides a computer-readable storage medium, on which a video-based classification label marking processing program is stored, and when the video-based classification label marking processing program is executed by a processor, the step of any one of the video-based classification label marking processing methods is implemented.

[0039] As can be seen from the above, the present invention discloses a method, device, terminal and medium for processing classification label marks based on videos. The present invention is mainly used for automatically labeling unlabeled videos. The present invention collects user data that actively acts on a certain video, filters out the video labels that the user has watched recently, and obtains high-quality video labels that the user has recently watched after data cleaning of these video labels. Different weights are assigned according to the user's completion rate of a certain video, whether to like it, and whether to comment on it. Based on the clustering results of the weighted video labels as video labels, the accuracy of the video labels is evaluated, and the weights and label cleaning weights are adjusted feedback until a more ideal effect is achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0041] Figure 1It is a flowchart of a video-based classification label processing method provided by an embodiment of the present invention.

[0042] Figure 2 The present invention is implemented Figure 1 Specific flow diagram of step S100;

[0043] Figure 3 The present invention is implemented Figure 1 Specific flow diagram of step S200;

[0044] Figure 4 The present invention is implemented Figure 1 Specific flow diagram of step S300;

[0045] Figure 5 The present invention is implemented Figure 1 Specific flow diagram of step S400;

[0046] Figure 6 It is a specific flow chart of a specific embodiment of the present invention using a video-based classification label identification processing method.

[0047] Figure 7 It is a structural schematic diagram of a video-based classification label marking processing device provided by an embodiment of the present invention.

[0048] Figure 8 It is a block diagram of the internal structure principle of an intelligent terminal provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0049] In the following description, specific details such as specific system structures, technologies, etc. are provided for the purpose of illustration rather than limitation, so as to provide a thorough understanding of the embodiments of the present invention. However, it should be clear to those skilled in the art that the present invention may be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to prevent unnecessary details from obstructing the description of the present invention.

[0050] It should be understood that when used in this specification and the appended claims, the term "comprising" indicates the presence of described features, integers, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or groups thereof.

[0051] It should also be understood that the terms used in the present specification are only for the purpose of describing specific embodiments and are not intended to limit the present invention. As used in the present specification and the appended claims, unless the context clearly indicates otherwise, the singular forms "a", "an" and "the" are intended to include plural forms.

[0052] It should be further understood that the term "and / or" used in the present description and the appended claims refers to any and all possible combinations of one or more of the associated listed items, and includes these combinations.

[0053] As used in this specification and the appended claims, the term "if" may be interpreted as "when" or "upon" or "in response to determining" or "in response to detecting," depending on the context. Similarly, the phrases "if it is determined" or "if [described condition or event] is detected" may be interpreted as meaning "upon determination" or "in response to determining" or "upon detection of [described condition or event]" or "in response to detecting [described condition or event]," depending on the context.

[0054] The following is a clear and complete description of the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0055] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0056] With the development of science and technology, especially the rapid development of Internet technology, short videos and mini video programs are increasingly occupying people's spiritual life. Video uploaders can record their own lives or shoot other things they are interested in and want to express. Video viewers can also get a happy video experience by using fragmented time, making their short fragmented time meaningful. However, with the surge in the number of short videos and mini videos, how video platforms or websites manage the content of mini videos and the classification and labeling of mini videos has become a major difficulty.

[0057] Video tags in the prior art are mostly keywords related to the video selected by the user, but the video platform and the user itself cannot guarantee whether the video keywords or tags selected by the user are truly consistent with the video content. Therefore, an intelligent method for selecting tags for user videos is extremely important.

[0058] In order to solve the problems of the prior art, the solution of the present invention collects user data that actively engages in a certain video, filters out the video tags that the user has watched recently, and obtains high-quality video tags that the user has recently watched after data cleaning of these video tags. Different weights are assigned according to the user's completion rate of a certain video, whether they like it, and whether they comment on it. Based on the clustering results of the weighted video tags as video tags, the accuracy of the video tags is evaluated, and the weights and the weights of the tag cleaning are adjusted in a feedback manner until a more ideal effect is achieved.

[0059] Exemplary Methods

[0060] like Figure 1 As shown, the embodiment of the present invention provides a video-based classification label processing method. Specifically, the method includes the following steps:

[0061] Step S100, collecting and obtaining active viewing behavior data of users who have active viewing behavior for a specified video;

[0062] In this embodiment, the short video platform collects active viewing behavior data of users who watch a specified video such as "A Bite of China" and have active viewing behavior. The active viewing behavior of the user can be determined by quantified data, including parameters that can actually reflect user behavior, such as the number of views or openings, viewing time, whether to participate in comments, whether to like, etc., and different weights are given according to different user behaviors. For example, compared with the number of openings, likes and comments can better reflect the user's active viewing of the video, so the weights of likes and comments are higher and the weights corresponding to the number of views or openings are lower. At this time, the platform collects the active viewing behavior data of all users who watch the video "A Bite of China". Through the steps of this method, the degree of interest of all users when watching a certain video can be obtained, and when the active viewing behavior data is high, it can be considered that the corresponding user has a greater interest in the video, and when the active viewing behavior data is low, it is considered that the corresponding user has little interest in the video, which is conducive to screening out user groups who really like and are interested in a certain video.

[0063] Step S200: based on the active viewing behavior data, confirm the user group data whose liking degree for the specified video is greater than a predetermined value, and filter out the first video data watched by the user group data in the most recent specified time period;

[0064] In this embodiment, the platform confirms the user group data whose liking degree for the specified video is greater than a predetermined value based on the above-obtained active viewing behavior data. The predetermined value is a pre-set value for the active viewing behavior data. Through the predetermined value, it is possible to determine which user likes the video and which user does not like the video by judging the above-obtained active viewing behavior data of all users. And the platform selects the users who like the video according to the judgment result, and obtains the data of the users who like the video to obtain the user group data whose liking degree for the specified video is greater than the predetermined value. Through this step, it is possible to determine which part of all users who watch a certain video are users who really like the video, and obtain the user data of all users who like the video.

[0065] Furthermore, the platform further extracts the first video data watched by all users of the user group data within a specified time period based on the filtered user group data. The staff of the platform can control the amount of video data obtained by setting the length of the specified time period, which is conducive to customizing the amount of computing data according to the computing power of the platform server, and is conducive to truly obtaining the types of videos that users like within a fixed time period according to actual conditions.

[0066] Step S300: Based on the filtered first video data, data cleaning is performed on the first video data to filter out second video data whose user preference is greater than a predetermined value, and a video type label watched by the user group data in a recently specified time period is obtained;

[0067] In this embodiment, the platform filters the first video data based on the filtered first video data, and filters out the second video data whose user's favorite degree is greater than a predetermined value. The predetermined value here and the predetermined value in step 100 can be a predetermined value or another value. The predetermined value is used to determine which videos in the first video data are opened by the user but actually disliked, and by cleaning out these videos that the user has opened but actually disliked, only the second video data of the videos that the user in the user group data shows as liking to watch is left, and further by obtaining the video type label of the second video data, the video type label watched by the user group data in the most recent specified time period is obtained. Through the steps of this method, the favorite videos watched by the user within the specified time are accurately obtained, and the video type label corresponding to the favorite video of the user within the specified time is further obtained, which serves as a basis for the subsequent judgment of the specified video, such as "A Bite of China", and effectively reduces the judgment error.

[0068] Step S400: Based on the obtained video type labels of the user group data watched in the most recent specified time period, a predetermined number of type labels are selected from the type labels of the specified video by different user groups as the final video labels of the specified video, and the label weights owned by the corresponding users are assigned according to the user's preference.

[0069] In this embodiment, the platform obtains a preset number of video type tags corresponding to multiple different user groups as the final video tags of the designated video based on the video type tags corresponding to the favorite videos of the users within the specified time. The multiple different user groups are several user groups divided according to user characteristics in the user group data, and the characteristics include age, gender, income, etc. In this step, by redividing the user group data, the video tags of the designated video can be obtained from multiple user perspectives, making the video tags of the designated video more diversified and more in line with the preferences of all users, and the designated video can be more accurately labeled and pushed to the target user group, indirectly improving the intelligent service and user stickiness of the short video software platform.

[0070] As can be seen from the above, the video-based label processing method provided by the embodiment of the present invention collects and obtains the user viewing behavior data of the specified video, and counts the user's preference for the specified video; based on the preference, multiple users whose preference is greater than a predetermined value are screened out to form user group data, and the video data watched by the user group data at the most recent specified time are counted; the video data watched by each user in the user group data at the most recent specified time is cleaned, and the video data whose preference is greater than a predetermined value is screened out from the video data watched by each user at the most recent specified time, and the corresponding video type label is screened out; based on the corresponding video type label screened out, according to the number of occurrences and the user's preference, the K video labels with the highest sum of weights are obtained as the most recently watched video labels of the corresponding user; based on the video type labels watched by the user group data at the most recent specified time period, a predetermined number of type labels are selected from the type labels of the specified video by different user groups as the final video labels of the specified video, and the label weights owned by the corresponding user are assigned according to the user's preference. Compared with the prior art, the video label of the specified video can be obtained by intelligent analysis, avoiding the problem of manually adding video labels and the inability of video platform auditors to define video labels and types.

[0071] Specifically, in this embodiment, a short video platform commonly used by users is taken as an example. When the above-mentioned short video platform is other video platforms, the specific scheme in this embodiment can be referred to.

[0072] In one application scenario, the short video platform mobile phone obtains the base viewing behavior data of users who have active viewing behavior for a specified video.

[0073] Specifically, Figure 2 As shown, the above step S100 includes:

[0074] Step S101, collecting and obtaining user active viewing behavior data for a specified video according to a user active viewing behavior indicator;

[0075] Step S102: Statistically analyzing the user's preference for the specified video based on the user's active viewing behavior data for the specified video.

[0076] For example, a user uploaded a vlog video of life, which is called video A. When uploading video A, the user did not manually fill in the video tags, but chose the option of adding tags intelligently on the short video platform. After the video is reviewed and released, the video will be randomly assigned to the recommended areas of users of different age groups, income situations, and genders. When each user opens the video A, the platform records the active viewing behavior indicators of all users, and collects the active viewing behavior data of all users who watch video A when watching video A. The active viewing behavior indicators include the number of views, viewing time, whether to comment, and whether to like. The corresponding active viewing behavior data are recorded as v1, v2, v3, and v4, respectively. Further, a weight is set for each active viewing behavior indicator, corresponding to p1, p2, p3, and p4, respectively. Then, through the above-mentioned active viewing behavior data and its weight, each user's liking for the video A can be obtained L=v1p1+v2p2+v3p3+v4p4. In this step, the viewing behavior of each user is converted into quantitative data to learn how much each user likes video A, and then the audience, potential, video tags and other information of the video can be obtained based on the distribution of the likes of each user or the characteristics of the users who like the video.

[0077] In one application scenario, the platform determines a user group whose degree of liking for a specified video is greater than a preset value based on the user's active viewing behavior data, obtains data of the corresponding user group, and obtains the first video data corresponding to the video watched within a specified time in the user group data.

[0078] Specifically, Figure 3 As shown, the above step S200 includes:

[0079] Step S201, based on the user preference of each user confirmed by the active viewing behavior data for the specified video, confirm the user group data whose preference for the specified video is greater than a predetermined value;

[0080] Step S202: based on the user group data, filter out the first video data watched by the user group data in a recent specified time period.

[0081] For example, the short video platform uses the above formula L=v1p1+v2p2+v3p3+v4p4 to calculate the degree of each user's liking for video A based on the determined active viewing behavior data of all viewers. And all users are screened according to the predetermined value L1 related to the degree of liking, for example, only users whose values ​​after the calculation of the degree of liking L exceed the predetermined value L1 are determined to be users who like video A, and all users who are judged to like video A, that is, users whose calculated degree of liking L is greater than the predetermined value L1, are set as user group data. By combining the user liking formula and the method of the predetermined value L1, only the user group data of users who are interested in video A is screened out, which is conducive to more accurate analysis of the video tag of video A in the subsequent steps.

[0082] Furthermore, based on the user group data, data of all videos watched by all users in the user group data within a specified time period are filtered out, that is, the first video data.

[0083] In an application scenario, the short video platform performs data cleaning and filtering on the first video data based on the filtered first video data, obtains only the second video data that the user likes in the first video data, and further obtains the video type label corresponding to the second video data.

[0084] Specifically, Figure 4 As shown, the above step S300 includes:

[0085] Step S301: Based on the first video data that the user group data has recently watched in a specified time period, data cleaning is performed on the first video data that each user in the user group data has recently watched in a specified time period, video data whose user preference is less than a predetermined value is cleaned out, and the existing tags of the remaining videos are counted;

[0086] Step S302, taking the k video tags with the highest number of occurrences as the video tags most recently watched by the corresponding user;

[0087] Step S303: Count the video tags most recently watched by each user in the user group data to obtain the video type tags watched by the user group data in the most recent specified time period.

[0088] For example, the short video platform compares and judges each video watched by each user in the first video data based on the first video data that has been screened, and calculates the degree of liking L of each user when watching each video through the calculation formula of the degree of liking L. When the degree of liking of a user when watching a certain video B is less than the predetermined value L1, the data of video B is deleted, that is, the user has watched video B but actually does not like watching video B; when the degree of liking of a user when watching video C is greater than the predetermined value L1, it is judged that the user likes video C, and the data of video C is retained. All the retained videos are integrated through the above method, and the tags of all videos liked by the user and the number of times the tags appear are counted.

[0089] Furthermore, the platform sorts the number of labels of the videos retained by each user, and uses the k video labels with the highest number of appearances as the corresponding user's most recently watched video labels. For example, in the videos watched by a child, the number of appearances of the video label of Honor of Kings is 20, the number of appearances of the funny video label is 18, and the number of appearances of the vlog video label is 15. When k is 3, the video labels of Honor of Kings, Funny, and Vlog are used as the video labels of the user's most recently watched video. Finally, the platform obtains the video labels of each user's most recently watched video according to the above steps, and counts them as the video type labels watched by the user group data in the most recent specified time period. Through this step, the video labels corresponding to the videos with high popularity among all users in the user group data are obtained, and the video labels preferred by users in the user group data are accurately obtained. By sorting and obtaining some video labels, it is avoided that the user accidentally opens a video but its positive behavior data meets the requirements and is counted as a favorite video, which causes errors when obtaining video labels.

[0090] In one application scenario, the short video platform selects a predetermined number of video type tags from the video tags corresponding to different user groups as the final video type tags for the specified video based on the video type tags watched by the acquired user group data within a specified time period, and assigns weights to the tags owned by the corresponding users according to the user's preference.

[0091] Specifically, Figure 5 As shown, the above step S400 includes:

[0092] Step S401: Based on the obtained video type labels of the user group data watched in the most recent specified time period, assign corresponding weights to the users according to the degree of their active behaviors towards the specified videos;

[0093] Step S402: clustering users according to the partial features to obtain type labels of the specified video for different user groups;

[0094] Step S403: Select a specified number of type tags from type tags for the specified video by different user groups as type tags for the specified video.

[0095] For example, the short video platform assigns corresponding weights to users based on the labels of the types of videos watched in the specified time period based on the obtained user group data, according to the degree of liking of the users for the positive behavior of the specified video. For example, when a user watches a specified video, the platform obtains the base viewing behavior data of the user, and the liking degree L value corresponding to the user is high, then the video type label of the user should be given a higher weight. Further, the users are clustered and classified according to their characteristics to obtain the type labels of the specified video for different user groups. For example, when the male user group pays attention to the sports car in the video when watching a certain video, the type label provided by the male user group to the specified video is sports car, while the female user group pays more attention to the cosmetics appearing in the video when watching the video, and the type label provided by the female user group to the specified video is beauty makeup. It is further limited that each user group can only contribute a fixed number of type labels to the specified video, and according to the actual situation, each user group can also be set to contribute type labels according to the proportion of the number of users, and the total number of type labels provided shall not exceed the preset number, such as 10.

[0096] In a more specific embodiment, step S400 further includes:

[0097] Extracting user features of users who are interested in the specified video, and performing unsupervised clustering to obtain different user groups G1, G2, Gn;

[0098] And assign different weights according to the number of different groups, w1, w2, wn;

[0099] Suppose whether a label appears in G1 is is_label1, where it is 1 if it appears and 0 if it does not appear. The final formula for calculating the label weight is: is_label1*w1+is_label2*w2+...+is_labeln*wn;

[0100] The r video tags with the largest number of tag appearances are used as the video tags of the specified video.

[0101] For example, the short video platform extracts the user characteristics of users who are interested in the specified video, that is, the users in the user group data, and performs unsupervised clustering to obtain different user groups G1, G2...Gn. The user characteristics include the age group, income, gender and other data of the user account. The unsupervised clustering is a modeling framework in deep learning, where the users in the user group data can be classified according to user characteristics. Further, different weights w1, w2...wn are assigned according to the number of users in different groups. When calculating the statistical label weight, a variable is_label1 is set to define whether the parameter appears. When it appears, the variable is 1, and when it does not appear, the variable is 0. The final formula for statistical label weight is: is_label1*w1+is_label2*w2+...+is_labeln*wn. The r video labels with the most label appearances are used as the video labels of the specified video. In this step, by dispersing the users in the user group data into multiple different user groups, obtaining the video tags with the highest number of appearances or the highest weights in each user group as the video tags of the specified video, it is possible to collect tags that various users are interested in when facing the video, so that after the video obtains the tag through the intelligent setting, the specified video can be pushed to all target users to the greatest extent, thereby improving the intelligence of the short video platform and the user stickiness of users.

[0102] After step S400, the method further includes:

[0103] The accuracy of video labels is evaluated, and the weights and label cleaning weights are adjusted based on feedback.

[0104] For example, videos with intelligently assigned video tags are placed in the target user's video recommendation area, and the video views, completion rate and other parameters are used to provide feedback on whether the video tags are qualified. The method of intelligently assigning tags on the short video platform is adjusted by adjusting the weights, predetermined values ​​and clustering features in the above method steps. A more intelligent and successful intelligent tag identification process is achieved, providing users with more accurate video push and reducing the workload of video platform reviewers. In addition, compared with manually assigned tags, video tags derived from intelligent judgment of user viewing data can better stand from the user's perspective to think about how to attract the user's attention and intelligently push, which can increase video exposure compared to manual labeling methods.

[0105] The present invention is further described in detail below by using specific application examples:

[0106] like Figure 6 As shown, the video-based classification label identification processing method described in the specific embodiment of the present invention includes the following steps:

[0107] Step S10, start, proceed to step S11;

[0108] Step S11, collecting user data of users who watch the video actively during cold start, and proceeding to step S12;

[0109] Step S12: Collect the video data that the user has recently watched, and proceed to step S13

[0110] Step S13: Clean the data to obtain the type label of the video that the user has recently watched, and then proceed to step S14;

[0111] Step S14, assigning corresponding weights to users according to the degree of their positive behaviors in the video, and proceeding to step S15;

[0112] Step S15, clustering users according to some features to obtain type labels of the video for different user groups, and then proceeding to step S16;

[0113] Step S16, selecting a suitable number of type tags from different user groups as the tags of the video, and proceeding to step S17;

[0114] Step S17, testing the accuracy of the video tag to see if it is qualified, if it is qualified, proceeding to step S20, if it is not qualified, returning to step S14 or step S16 to adjust the relevant parameters according to the actual situation;

[0115] Step S20, end.

[0116] As can be seen from the above, in a specific application embodiment of the present invention, the video platform collects user data of users who have cold-started and actively watched a certain video. In this embodiment, the cold-start refers to a video that is randomly pushed to the video user when the user opens it without a video tag. Further, the data of the videos recently watched by this part of users is obtained, and the data of the recently watched videos is cleaned, and only the data of the videos in which the user shows active behavior when watching is retained, such as the user likes, recommends or watches the video multiple times, etc. Through this step, the data of the video that the user accidentally opens or does not like after opening is cleared. Further, the user is given a corresponding weight according to the degree of the user's active behavior towards the video. For example, user A only watches the video completely, and user B not only watches the video completely but also likes, comments and shares the video, indicating that user B is more likely to like this type of video than user A. Then, the video tag in the video liked by user B has a greater probability of reflecting the video type tag of the video, so the user B is set to have a greater weight. The platform clusters users based on their characteristics, obtains type labels for the video from different user groups, and selects an appropriate number of type labels from the type labels for the video from different user groups as labels for the video. For example, when all users watching a video are boys, the obtained video type labels have heavier boys' characteristics, and the labels of the boys' group with a larger base will fully represent the type labels of the video, which will result in the video being pushed to the male user group only based on the labels with boys' characteristics after being labeled, and cannot be pushed to different user groups in many aspects, such as user groups of different genders, user groups with different incomes, and user groups of different age groups. Each user group has different views and labels for the same video. Through this step, the platform can give the video more comprehensive and diversified video type labels. Finally, the platform will deliver and recommend the well-tagged videos to target users, and record feedback information to determine whether the video has achieved a good video promotion effect based on the current tag situation. If so, the process ends; if not, the process returns to step S14 or step S16, and adjusts the weights of different users and the number of video tags selected by different user groups or other parameters to adjust the tags of the video until the video achieves the promotion effect pre-set by the platform.

[0117] Exemplary Devices

[0118] like Figure 7 As shown in , corresponding to the above-mentioned video-based classification label mark processing method, an embodiment of the present invention further provides a video-based classification label mark processing device, and the above-mentioned video-based classification label mark processing device includes:

[0119] The viewing behavior data acquisition module 710 is used to collect and acquire active viewing behavior data of users who have active viewing behavior for a specified video;

[0120] A screening module 720 is used to identify user group data whose liking degree for the specified video is greater than a predetermined value based on the active viewing behavior data, and screen out the first video data watched by the user group data in the most recent specified time period;

[0121] The video type label initial confirmation module 730 is used to perform data cleaning on the first video data to filter out the second video data whose user preference is greater than a predetermined value based on the filtered first video data, and obtain the video type label watched by the user group data in the most recent specified time period;

[0122] The final video tag confirmation module 740 is used to select a predetermined number of type tags from the type tags of the specified video by different user groups based on the video type tags watched by the user group data in the most recent specified time period as the final video tags of the specified video, and assign weights to the tags owned by the corresponding users according to the user's preference.

[0123] Specifically, in this embodiment, the specific functions of each module of the above-mentioned video-based classification label mark processing device can refer to the corresponding description in the above-mentioned video-based classification label mark processing method, and will not be repeated here.

[0124] Based on the above embodiments, the present invention further provides an intelligent terminal, whose principle block diagram can be shown as follows: Figure 8 As shown. The above-mentioned intelligent terminal includes a processor, a memory, and a network interface connected through a system bus. Among them, the processor of the intelligent terminal is used to provide computing and control capabilities. The memory of the intelligent terminal includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a video-based classification label mark processing program. The internal memory provides an environment for the operation of the operating system and the video-based classification label mark processing program in the non-volatile storage medium. The network interface of the intelligent terminal is used to communicate with an external terminal through a network connection. When the video-based classification label mark processing program is executed by the processor, the steps of any one of the above-mentioned video-based classification label mark processing methods are implemented.

[0125] Those skilled in the art will understand that Figure 8 The principle block diagram shown in the figure is only a block diagram of a partial structure related to the scheme of the present invention, and does not constitute a limitation on the smart terminal to which the scheme of the present invention is applied. The specific smart terminal may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0126] In one embodiment, a smart terminal is provided, the smart terminal comprising a memory, a processor, and a video-based classification label mark processing program stored in the memory and executable on the processor, the video-based classification label mark processing program performing the following operation instructions when executed by the processor:

[0127] A. Collect and obtain active viewing behavior data of users who have active viewing behavior for a specified video;

[0128] B. Based on the active viewing behavior data, confirm the user group data whose liking degree for the specified video is greater than a predetermined value, and filter out the first video data watched by the user group data in the most recent specified time period;

[0129] C. Based on the filtered first video data, the first video data is cleaned to filter out second video data whose user preference is greater than a predetermined value, and a video type label watched by the user group data in a recently specified time period is obtained;

[0130] D. Based on the obtained video type labels of the user group data watched in the most recent specified time period, a predetermined number of type labels are selected from the type labels of the specified video by different user groups as the final video labels of the specified video, and the label weights owned by the corresponding users are assigned according to the user's preference.

[0131] An embodiment of the present invention also provides a computer-readable storage medium, on which a video-based classification label marking processing program is stored. When the video-based classification label marking processing program is executed by a processor, the steps of any one of the video-based classification label marking processing methods provided by the embodiment of the present invention are implemented.

[0132] In summary, the present invention discloses a video-based classification label mark processing method, device, intelligent terminal and computer-readable storage medium, wherein the method includes: collecting and obtaining active viewing behavior data of users who have active viewing behavior for a specified video; based on the active viewing behavior data, confirming the user group data whose liking degree for the specified video is greater than a predetermined value, and filtering out the first video data watched by the user group data in the most recent specified time period; based on the filtered first video data, performing data cleaning on the first video data to filter out the second video data whose liking degree for the user is greater than a predetermined value, and obtaining the video type label watched by the user group data in the most recent specified time period; based on the obtained video type label watched by the user group data in the most recent specified time period, selecting a predetermined number of type labels from the type labels of the specified video of different user groups as the final video label of the specified video, and assigning the label weight owned by the corresponding user according to the user's liking degree. Compared with the prior art, the present invention can avoid the problem of poor video push effect caused by incorrect user upload labels, and can also help video platforms to quickly and accurately analyze, process and classify massive video resources, improve the speed of automatic labeling, greatly save manual review efficiency, and save costs.

[0133] It should be understood that the serial numbers of the steps in the above embodiments do not imply a sequence of execution, and the execution sequence of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0134] Those skilled in the art can clearly understand that, for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example for illustration. In practical applications, the above-mentioned function allocation can be completed by different functional units and modules as needed, that is, the internal structure of the above-mentioned device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiment can be integrated in a processing unit, or each unit can exist physically separately, or two or more units can be integrated in one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of software functional units. In addition, the specific names of the functional units and modules are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of the present invention. The specific working process of the units and modules in the above-mentioned system can refer to the corresponding process in the aforementioned method embodiment, which will not be repeated here.

[0135] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described or recorded in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0136] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0137] In the embodiments provided by the present invention, it should be understood that the disclosed apparatus / terminal equipment and method can be implemented in other ways. For example, the apparatus / terminal equipment embodiments described above are only illustrative, for example, the division of the above modules or units is only a logical function division, and in actual implementation, other division methods can be used, for example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed.

[0138] If the above-mentioned integrated module / unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such an understanding, the present invention implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The above-mentioned computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of the above-mentioned various method embodiments when executed by the processor. Among them, the above-mentioned computer program includes computer program code, and the above-mentioned computer program code can be in source code form, object code form, executable file or some intermediate form. The above-mentioned computer-readable medium may include: any entity or device capable of carrying the above-mentioned computer program code, recording medium, U disk, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electric carrier signal, telecommunication signal and software distribution medium. It should be noted that the content contained in the above-mentioned computer-readable storage medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction.

[0139] The embodiments described above are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, it should be understood by those skilled in the art that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features therein may be replaced by equivalents. However, these modifications or replacements do not deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should be included in the protection scope of the present invention.

Claims

1. A video-based classification label processing method, characterized in that: The method comprises: Collect and obtain active viewing behavior data of users who have active viewing behavior for a specified video; Based on the active viewing behavior data, confirm the user group data whose liking degree for the specified video is greater than a predetermined value, and filter out the first video data watched by the user group data in the most recent specified time period; Based on the filtered first video data, data cleaning is performed on the first video data to filter out second video data whose user preference is greater than a predetermined value, and a video type label watched by the user group data in a recently specified time period is obtained; Based on the obtained video type tags watched by the user group data in the most recent specified time period, a predetermined number of type tags are selected from the type tags of the specified video by different user groups as the final video tags of the specified video, and a weight of the tag owned by the corresponding user is assigned according to the user's preference; The step of selecting a predetermined number of type tags from the type tags of the specified video by different user groups based on the obtained video type tags watched in the most recent specified time period as the final video tags of the specified video, and assigning weights to the tags owned by the corresponding users according to the user's preference includes: Extracting user features of users who are interested in the specified video, and performing unsupervised clustering to obtain different user groups G1, G2, Gn; And assign different weights according to the number of different groups, w1, w2, wn; Suppose whether a label appears in G1 is is_label1, where it is 1 if it appears and 0 if it does not appear. The final formula for calculating the label weight is: is_label1*w1+is_label2*w2+...+is_labeln*wn; The r video tags with the largest number of tag appearances are used as the video tags of the specified video; The step of selecting a predetermined number of type tags from the type tags of the specified video by different user groups as the final video tags of the specified video based on the obtained type tags of the video watched by the user group data in the most recent specified time period, and assigning weights to the tags owned by the corresponding users according to the user's preference degree, comprises: The accuracy of video labels is evaluated, and the weights and label cleaning weights are adjusted based on feedback.

2. The video-based classification label processing method according to claim 1 is characterized in that: The step of collecting and obtaining active viewing behavior data of users who have active viewing behavior for a specified video includes: According to the user's active viewing behavior indicators, collect and obtain the user's active viewing behavior data for the specified video; Based on the user's active viewing behavior data for the specified video, a statistical analysis is performed on the user's liking for the specified video.

3. The video-based classification label processing method according to claim 1, characterized in that: The step of confirming user group data whose liking degree for the specified video is greater than a predetermined value based on the active viewing behavior data, and filtering out the first video data watched by the user group data in the most recent specified time period includes: Based on the user preference of each user for the specified video confirmed by the active viewing behavior data, confirm the user group data whose preference for the specified video is greater than a predetermined value; Based on the user group data, first video data watched by the user group data in a recent specified time period is filtered out.

4. The video-based classification label processing method according to claim 1 is characterized in that: The step of performing data cleaning on the first video data to filter out second video data whose user preference is greater than a predetermined value based on the filtered first video data, and obtaining a video type label watched by the user group data in a recently specified time period comprises: Based on the first video data that the filtered user group data has watched in the most recent specified time period, data cleaning is performed on the first video data that each user in the user group data has watched in the most recent specified time period, video data whose user preference is less than a predetermined value is cleaned out, and the existing tags of the remaining videos are counted; The k video tags with the highest number of occurrences are used as the video tags most recently watched by the corresponding user; And count the video tags recently watched by each user in the user group data to obtain the video type tags watched by the user group data in the most recent specified time period.

5. The video-based classification label processing method according to claim 1, characterized in that: The step of selecting a predetermined number of type tags from the type tags of the specified video by different user groups based on the obtained video type tags watched in the most recent specified time period as the final video tags of the specified video, and assigning weights to the tags owned by the corresponding users according to the user's preference includes: Based on the obtained video type labels of the user group data watched in the most recent specified time period, assigning corresponding weights to the users according to the degree of their positive behavior towards the specified videos; Clustering users according to user characteristics in user group data to obtain type labels for the specified video by different user groups; A specified number of type tags are selected from type tags of the specified video by different user groups as type tags of the specified video.

6. A video-based classification label processing device, characterized in that: The device comprises: A viewing behavior data acquisition module is used to collect and acquire active viewing behavior data of users who have active viewing behavior for a specified video; A screening module, for identifying user group data whose liking for the specified video is greater than a predetermined value based on the active viewing behavior data, and screening out first video data watched by the user group data in a recent specified time period; A video type label initial confirmation module is used to perform data cleaning on the first video data to filter out second video data whose user preference is greater than a predetermined value based on the first video data filtered out, and obtain the video type label watched by the user group data in the most recent specified time period; A final video tag confirmation module is used to select a predetermined number of type tags from the type tags of the specified video by different user groups based on the type tags of the videos watched in the most recent specified time period by the user group data as the final video tags of the specified video, and assign a weight to the tag owned by the corresponding user according to the user's preference; The step of selecting a predetermined number of type tags from the type tags of the specified video by different user groups based on the obtained video type tags watched in the most recent specified time period as the final video tags of the specified video, and assigning weights to the tags owned by the corresponding users according to the user's preference includes: Extracting user features of users who are interested in the specified video, and performing unsupervised clustering to obtain different user groups G1, G2, Gn; And assign different weights according to the number of different groups, w1, w2, wn; Suppose whether a label appears in G1 is is_label1, where it is 1 if it appears and 0 if it does not appear. The final formula for calculating the label weight is: is_label1*w1+is_label2*w2+...+is_labeln*wn; The r video tags with the largest number of tag appearances are used as the video tags of the specified video; The step of selecting a predetermined number of type tags from the type tags of the specified video by different user groups as the final video tags of the specified video based on the obtained type tags of the video watched by the user group data in the most recent specified time period, and assigning weights to the tags owned by the corresponding users according to the user's preference degree, comprises: The accuracy of video labels is evaluated, and the weights and label cleaning weights are adjusted based on feedback.

7. An intelligent terminal, characterized in that: The intelligent terminal includes a memory, a processor, and a video-based classification label marking processing program stored in the memory and executable on the processor. When the video-based classification label marking processing program is executed by the processor, the steps of the video-based classification label marking processing method as described in any one of claims 1 to 5 are implemented.

8. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a video-based classification label marking processing program, and when the video-based classification label marking processing program is executed by a processor, the steps of the video-based classification label marking processing method as described in any one of claims 1-5 are implemented.

Images