WO2020259572A1 - 用于负反馈的标签确定方法、视频推荐方法、装置、设备和存储介质 - Google Patents
用于负反馈的标签确定方法、视频推荐方法、装置、设备和存储介质 Download PDFInfo
- Publication number
- WO2020259572A1 WO2020259572A1 PCT/CN2020/098065 CN2020098065W WO2020259572A1 WO 2020259572 A1 WO2020259572 A1 WO 2020259572A1 CN 2020098065 W CN2020098065 W CN 2020098065W WO 2020259572 A1 WO2020259572 A1 WO 2020259572A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- tag
- original
- video data
- user
- video
- Prior art date
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/70—Information retrieval; Database structures therefor; File system structures therefor of video data
- G06F16/73—Querying
- G06F16/735—Filtering based on additional data, e.g. user or group profiles
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/70—Information retrieval; Database structures therefor; File system structures therefor of video data
- G06F16/78—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually
- G06F16/7867—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using information manually generated, e.g. tags, keywords, comments, title and artist information, manually generated time, location and usage information, user ratings
Definitions
- the embodiments of the present application relate to the technical field of video recommendation, for example, to a label determination method for negative feedback, a video recommendation method, a label determination device for negative feedback, a video recommendation device, equipment, and storage medium.
- the video recommendation algorithm mainly recommends videos to the user based on the user's feedback.
- the user can provide negative feedback for the video, so as to collect the reasons why the user is not interested in the video to optimize the recommendation algorithm.
- a negative feedback method in the related art is shown in Figure 1a and Figure 1b. This method sets a negative feedback button that expresses dislike or disinterest. The user uses the negative feedback button to give negative feedback to the video. Another negative feedback The method is shown in Figure 1c. After the negative feedback is awakened, multiple negative feedback options including dislike of the author, boring video, and content repetition are displayed for users as negative feedback choices.
- the embodiments of the present application provide a method for determining a label for negative feedback, a method for video recommendation, a device for determining a label for negative feedback, a video recommendation device, equipment, and storage medium, so as to solve the problem of negative feedback on video in related technologies.
- negative feedback has poor directivity, which leads to the problem that a relatively large proportion of videos in recommended videos make users disgusted after optimizing the recommendation algorithm based on negative feedback.
- the embodiment of the present application provides a method for determining a label for negative feedback, including:
- a candidate label is determined from the multiple original labels according to the probability that each original label is used for negative feedback.
- the embodiment of the present application provides a video recommendation method, including:
- the target label is a label selected from candidate labels when the user performs a negative feedback operation
- the candidate label is determined by the label determination method for negative feedback described in the embodiment of the present application.
- the embodiment of the present application provides a label determination device for negative feedback, including:
- the original tag acquisition module is set to acquire multiple original tags of video data
- the original tag parameter obtaining module is set to obtain the saliency value of each original tag and the user's preference for each original tag, wherein the saliency value of each original tag indicates that the video data with each original tag is in The proportion of all video data;
- a probability calculation module configured to calculate the probability that each original tag is used for negative feedback according to the significance value of each original tag and the user's preference for each original tag
- the candidate label determination module is configured to determine a candidate label from the multiple original labels according to the probability that each original label is used for negative feedback.
- An embodiment of the present application provides a video recommendation device, including:
- the user determination module is set to determine the user of the video to be recommended
- the video data acquisition module is set to acquire multiple video data
- a user tag determination module configured to determine the target tag of the user
- a target video data determining module configured to determine target video data from the plurality of video data based on the target tag
- the video push module is configured to push the target video data to the client terminal
- the target tag is a tag selected from candidate tags when the user performs a negative feedback operation, and the candidate tag is determined by the tag determination device for negative feedback described in this embodiment of the application.
- An embodiment of the present application provides a device, and the device includes:
- One or more processors are One or more processors;
- Storage device set to store one or more programs
- the one or more processors implement the label determination method and/or video for negative feedback according to any embodiment of the present application Recommended method.
- An embodiment of the present application provides a computer-readable storage medium on which a computer program is stored.
- the program When the program is executed by a processor, it implements the label determination method and/or video for negative feedback described in any embodiment of the present application. Recommended method.
- Figure 1a is a first schematic diagram of a user interface for negative feedback on video in related technologies
- Figure 1b is a second schematic diagram of a user interface for negative feedback on video in related technologies
- Figure 1c is a third schematic diagram of a user interface for negative feedback on video in related technologies
- FIG. 2 is a flowchart of a method for determining a label for negative feedback according to Embodiment 1 of the present application;
- 3A is a flowchart of a method for determining a label for negative feedback according to Embodiment 2 of the present application;
- FIG. 3B is a schematic diagram of the generation mechanism of the candidate tag of the video in the second embodiment of the present application.
- 3C is a schematic diagram of a user interface for negative feedback on a video provided in the second embodiment of the present application.
- FIG. 4 is a flowchart of a video recommendation method provided by Embodiment 3 of the present application.
- FIG. 5 is a structural block diagram of a label determining device for negative feedback provided in the fourth embodiment of the present application.
- FIG. 6 is a structural block diagram of a video recommendation device provided by Embodiment 5 of the present application.
- FIG. 7 is a structural block diagram of a device provided in Embodiment 6 of the present application.
- FIG. 2 is a flowchart of a method for determining a label for negative feedback according to the first embodiment of the application.
- the embodiment of the present application is applicable to the case of performing negative feedback on video data based on the label.
- the method can be used for negative feedback.
- the label determination device is implemented by means of software and/or hardware, and is integrated in the device that executes the method. As shown in Figure 2, the method may include the following steps:
- the original tag can be the keywords of the video data, the classification information of the video, etc. In practical applications, it can be a video data in advance based on crowdsourced manual annotation, visual automatic annotation, and comment-based automatic annotation when offline. Annotate tags. When the video data is played, multiple tags that have been annotated for the video data when offline can be obtained as the original tags.
- the original tag of the video data played by the user may be obtained.
- the video playback platform may be a variety of short video playback platforms, and the playback video data may be that after the user selects a video on the short video playback platform, the video data is played on the playback page.
- the saliency value of the tag expresses the degree of discrimination of the tag from the video content.
- the saliency value can be the coverage rate of the tag in all video data. For example, for a tag "student", in all video data on the entire video playback platform, more than 80% of the video data have the tag “student” ", the coverage rate of the label “student” is 80%, indicating that the label "student” is not significant. The label “student” has a low degree of distinguishing video data.
- the preference of a tag expresses the user’s degree of interest in the video data with the tag.
- For the video data with a tag if the user has played the video data completely, or executed the video data after playing the video data follow, share, comment and other positive operation behaviors indicate that the user is interested in the video data.
- the higher the proportion of video data with positive operation behavior indicates that the user has the tag The higher the preference of, the more interested the user is in the video data with this tag.
- the saliency value of each tag can be calculated offline, and for each user, the user's preference for each tag can be calculated based on the user's historical behavior data.
- S130 Calculate the probability that each original tag is used for negative feedback according to the significance value of each original tag and the user's preference for each original tag.
- the probability that the original tag is used for negative feedback is negatively related to the saliency value and preference degree, that is, the smaller the saliency value and preference degree, the higher the probability.
- the sum value of the saliency value and the preference degree of the original tag can be calculated, and then the difference between 1 and the sum value can be calculated as the probability of the original tag being used for negative feedback. Since the saliency value expresses the degree of discrimination of the tag from the video content, the preference degree expresses the user's degree of interest in the video, that is, the user's viewing intention.
- the tag is used as a negative feedback tag, the greater the probability of the tag, the Negative feedback information has a greater probability to reflect that users are not interested or even disgusted with videos with this tag.
- S140 Determine candidate tags from the multiple original tags according to the probability that each original tag is used for negative feedback.
- N original tags with the highest probability can be selected as candidate tags, so that the user can select target tags from the candidate tags for negative feedback on the video data.
- the candidate label is determined from the original label of the video data
- it can show a Negative feedback page, and display N candidate tags on the negative feedback page, so that the user can select candidate tags from the N candidate tags for negative feedback on the video data
- the user selects candidate tags from the N candidate tags for negative feedback
- the candidate tag selected by the user is used as the target tag, and the target tag is associated with the user.
- recommending a video to a user you can first obtain the target tag associated with the user, and optimize the recommendation algorithm through the target tag. For example, you can exclude the video data with the target tag from all video data, thereby recommending the video data of interest to the user , To avoid recommending to users video data containing target tags, causing users to be disinterested or even disgusted.
- This embodiment of the application obtains multiple original tags of video data, obtains the saliency value of each original tag and obtains the user's preference for each original tag, and uses the saliency and preference to calculate each original tag for negative feedback After the probability, the candidate label is determined from the multiple original labels according to the probability.
- the saliency value of the tag expresses the degree of differentiation of the tag to the video content
- the preference degree expresses the user's intention. Based on the saliency and preference degree, the probability of the tag being used for negative feedback is calculated, and the probability for negative feedback is determined according to the probability.
- Tags taking into account the video content and user intentions, the user selects the target tag from the candidate tags for negative feedback, and the feedback is highly directed. After optimizing the recommendation algorithm through the target tag selected by the user, it can reduce the number of users in the video recommended to the user. The proportion of disgusting videos improves the accuracy of video recommendations.
- FIG. 3A is a flowchart of a method for determining a label for negative feedback according to Embodiment 2 of this application. The embodiment of this application is described on the basis of Embodiment 1. As shown in FIG. 3A, the method may include the following steps :
- the number of video data contained in the video set V can be counted as the first video total.
- the video data with one tag can be found in the video set V through the association relationship between the tag and the video data, and then the number of video data with the tag can be counted as the total number of the second videos.
- S240 Calculate the ratio between the total number of the second videos and the total number of the first videos, and use the ratio as the saliency value of each original tag.
- the saliency value of a tag may be the coverage rate of the tag, that is, the proportion of video data with the tag in the video set V. Therefore, the saliency value of a tag can be represented by the ratio of the number of video data with the tag to the number of all video data. That is, the ratio of the total number of the second video and the total number of the first video is calculated as the saliency value of the tag. large, the lower the significance of a tag, significantly lower the value, the more significant the tag, for video tag set S T of each tag T i, can be calculated saliency value for each tag T i in real time in the above manner.
- the user data includes a video playlist and operation behavior data.
- the video playlist may be a list of video data played by the user, and the operation behavior of the user when the video data is played may be immediately after the user plays the video data. Close the video data, close the video data after playing the video data, comment on the video data, follow the video data, and share the video data.
- the video data requested by the user to be played may be determined, the video data is recorded in the video playlist, and the operation triggered by the user on the video data is recorded to generate the operation behavior data.
- the playlist V u contains video data played by the user. For example, when the user requests to play video data, the video data is recorded in the video playlist V u .
- the user's operation can be recorded to generate a play log as the operation behavior data.
- the play log records the start time of the user play video, the end time of the user play video, and the operations performed during the play.
- S260 Determine the user's preference for each original tag according to the user data.
- the video data with each original tag may be determined from the video playlist as the first video set corresponding to each original tag, and from the first video set corresponding to each original tag, Determine the video data with positive operation behavior data in a video set as the second video set corresponding to each original tag, and then determine the number of video data in the first video set corresponding to each original tag to obtain The total number of third videos corresponding to each original tag, and the number of video data in the second video set corresponding to each original tag are determined to obtain the total number of fourth videos corresponding to each original tag, and finally based on the The total number of fourth videos corresponding to each original tag and the total number of third videos corresponding to each original tag calculate the user's preference for each original tag, where the preference is positively correlated with the total number of fourth videos, and The total number of three videos is negatively correlated.
- the positive operation behavior can be the behavior that expresses the user's interest in a video data.
- the positive operation behavior can be at least one of playing the complete video, liking the video, sharing the video, following the user, and commenting on the video.
- the first video set with the tag T i can be determined from the video playlist V u played by the user, and the second video with the positive operation behavior can be determined from the first video set Then, the total number of video data in the first video set and the total number of video data in the second video set are used to calculate the preference.
- User can be calculated through the following label T i u of preference:
- M being a smoothing parameter
- a value of 30 can be seen from the above formula, for a tag T i, the video data has been played by the user having the label T i and the larger the number of video forward operation behavior, DESCRIPTION user bought the video data having the tag T i.
- the first system parameter and the second system parameter can be obtained first, the product of the first system parameter and the saliency value of each original tag is calculated to obtain the first product, and the second system parameter and the user's
- the second product is obtained by multiplying the preference degree of each original tag, the sum of the first product and the second product is calculated, and the difference between 1 and the sum is calculated as the probability that each original tag is used for negative feedback.
- ⁇ and ⁇ are system parameters with positive values.
- the system parameters can be adjusted according to user feedback after online testing.
- the original label Significant value of And preference The smaller the probability that it is used for negative feedback Bigger.
- the embodiment of the application calculates the probability of the original label for negative feedback based on the saliency value and the degree of preference, and determines the candidate label for negative feedback from the original label according to the probability, taking the video content and the user’s intention into consideration, and the user selects the candidate label from the Selecting the target tag to give negative feedback to the video data is highly directional. After optimizing the recommendation algorithm through the target tag selected by the user, it can reduce the proportion of videos recommended to the user that are offensive to the user and improve the accuracy of video recommendation .
- S280 Sort the multiple original tags in descending order according to the probability of each original tag being used for negative feedback.
- the original tags can be sorted according to the probability.
- the original tags can be sorted in descending order according to the probability, so as to determine the multiple original tags with the largest probability as candidate tags.
- the original tags are sorted in descending order according to the probability. The greater the probability, the greater the probability that when the original tag is used as a candidate tag for negative feedback, accurate negative feedback information will be collected.
- the first N original tags are used as candidate tags.
- FIG. 3B shows the candidate label generation mechanism of the currently played video data in an embodiment of the application, and the candidate label generation mechanism is as follows:
- the user When the user is playing a video, he obtains the m tags of the currently playing video, and obtains the user portrait generated offline.
- the user portrait records the n videos played by the user and the preference of the n tags of the n videos.
- obtain the saliency value of the m tags of the currently playing video that have been generated offline and then find the preference of the m tags of the currently playing video from the preference of n tags, and calculate the currently playing video through the preference and saliency
- the probability of each of the m tags used for negative feedback, and then the 3 tags with the highest probability are used as candidate tags, namely candidate tag 1, candidate tag 2, and candidate tag 3.
- the candidate tags are pushed to the client, and the client displays the negative feedback page.
- the negative feedback page displays 3 candidate tags for the user to select the target tag on the negative feedback page to give negative feedback to the video data.
- the user selects a candidate tag on the negative feedback page to give negative feedback to the video data, and the user
- the selected candidate tag is used as the user's target tag, so that when a video is recommended to the user, the recommendation algorithm is optimized according to the target tag to determine the video data to be recommended to the user.
- the value of the number N of candidate tags can be 3, that is, the three original tags with the highest probability are used as candidate tags. If it is detected that the user wakes up the negative feedback by long pressing the video playback interface, the negative feedback page can be popped up, and a total of 3 candidate tags of tag 1, tag 2 and tag 3 will be displayed on the negative feedback page.
- the candidate tag selected by the user is used as the user's target tag. For example, the user selects tag 1 from tag 1, tag 2 and tag 3, and tag 1 is the target tag.
- Tag 1 optimizes the recommendation algorithm to avoid recommending video data with tag 1 to users.
- the saliency value of the tag expresses the degree of differentiation of the tag to the video content
- the preference degree expresses the user's intention.
- the probability of the original tag being used for negative feedback is calculated, and the probability is determined to be used for negative feedback.
- the feedback candidate tags are displayed when the user's negative feedback operation on the video data is detected, and then based on the user's operation on the candidate tags, the target tags are determined from the candidate tags, and when the video is recommended to the user, the target tags are optimized
- a recommendation algorithm for users to recommend videos so that after the target tag is determined from the candidate tags for negative feedback, the negative feedback information obtained integrates the video content and user intent, and the negative feedback is highly directed.
- the recommendation algorithm is optimized by the target tag It can reduce the proportion of video data that is offensive to the user among the video data recommended to the user, and improve the accuracy of the video recommendation.
- Fig. 4 is a flowchart of a video recommendation method provided in the third embodiment of this application.
- the embodiment of this application is applicable to the situation of recommending videos to users.
- the method can be executed by a video recommendation device, which can use software and/ Or it can be implemented by hardware and integrated in the device that executes the method. As shown in Figure 4, the method may include the following steps:
- a video can be recommended to the user when a preset event is detected.
- the preset event can be the detection that the user logs in to the short video playback platform through the account, and the user performs page turning and refresh operations on the video preview interface, if When a preset event is detected, the user who triggers the preset event is the user of the video to be recommended.
- users can also be classified according to the user’s historical behavior, and a category of users is regarded as the user of the video to be recommended, for example, when there is a new
- video data if the video data is more suitable for elderly people to watch, users in the age range of 50-70 can be used as users of the video to be recommended.
- the embodiment of the present application does not limit the manner of determining the user of the video to be recommended.
- the video data that the user is interested in can be determined based on the user’s historical behavior data.
- the video data uploaded for the first time in a period of time can also be obtained, or the video data can be obtained randomly. limit.
- the target tag may be a tag selected from the candidate tags of the video data for negative feedback when the user performs negative feedback on the video data, that is, the target tag is the tag of the video data that the user is not interested or even disgusted with.
- Candidate labels can be determined in the following ways:
- S40 Determine candidate tags from the multiple original tags according to the probability that each original tag is used for negative feedback.
- the candidate tag is determined, when the user wakes up negative feedback on the video data, the candidate tag is pushed to the client to display the candidate tag on the negative feedback page of the client, and when the user receives the negative feedback page for the candidate tag During the selection operation of the user, the target tag for the user to negatively feedback the video data is determined from the displayed candidate tags according to the user's selection operation.
- Embodiment 1 For the steps of determining the candidate tag and the user's target tag, refer to Embodiment 1 and Embodiment 2, which will not be described in detail here.
- S340 Determine target video data from the multiple video data based on the target tag.
- those skilled in the art can also determine the video data that the user is not interested or even disgusted with from multiple video data according to the target tag in other ways, so as to avoid recommending to the user video data that the user is not interested in.
- This embodiment of the application The method of determining the target video data from multiple video data according to the target tag is not limited.
- the target video data can be pushed to the client to display the target video data on the client, and then recommend the target video data to the user.
- the client displays a recommendation list that includes multiple target videos Data for users to choose to play.
- the user's target tag is the candidate tag selected by the user during negative feedback.
- the candidate tag is determined based on the saliency value and preference of the candidate tag and the probability for negative feedback is calculated.
- the degree of discrimination and preference of content expresses the user's intention, so that after the user selects the target tag from the candidate tags, the negative feedback information obtained through the negative feedback of the target tag integrates the video content and the user's intention, and the negative feedback has high directivity.
- the target video data is determined from multiple video data through the target tag, which reduces the proportion of the video data that is offensive to the user in the target video data, and improves the accuracy of the video recommendation.
- FIG. 5 is a structural block diagram of a label determining device for negative feedback provided in the fourth embodiment of the present application.
- the label determining device for negative feedback in an embodiment of the present application may include the following modules: original tag The obtaining module 401 is set to obtain multiple original tags of the video data; the original tag parameter obtaining module 402 is set to obtain the salient value of each original tag and obtain the user's preference for each original tag, where each original tag The saliency value of the tag represents the proportion of the video data with each original tag in all video data; the probability calculation module 403 is set to be based on the saliency value of each original tag and the user’s contribution to each original tag.
- the preference degree of the tags calculates the probability that each original tag is used for negative feedback; the candidate tag determination module 404 is configured to determine a candidate tag from the multiple original tags according to the probability that each original tag is used for negative feedback.
- the original tag parameter acquisition module 402 includes: a first video total statistics sub-module, set to count the number of all video data to obtain the first video total; a second video total count sub-module, set to count the total number of videos The number of video data of the original tags is used to obtain the total number of second videos; the saliency calculation sub-module is configured to calculate the ratio between the total number of the second videos and the total number of the first videos, and use the ratio as the The significant value of the original label.
- the original tag parameter acquisition module 402 includes: a user data acquisition sub-module configured to acquire user data, the user data being used to indicate an operation triggered by the user when the video data is played; preferences;
- the degree determination sub-module is configured to determine the preference degree of the user for each original tag according to the user data.
- the user data includes a video playlist and operating behavior data
- the user data acquisition submodule includes: a video data determining unit configured to determine the video data requested by the user to be played; and a video data recording unit configured to The video data is recorded in a video playlist; the operation behavior data recording unit is configured to record operations triggered by the user on the video data to generate operation behavior data.
- the original tag parameter acquisition module 402 includes: a first video set determination sub-module configured to determine from the video playlist video data with each original tag as the corresponding The first video set; the second video set determining sub-module, configured to determine from the first video set corresponding to each original tag the video data with positive operation behavior data as each original tag Corresponding second video set; a third video total count sub-module, configured to determine the number of video data in the first video set corresponding to each original tag, and obtain the third total number of videos corresponding to each original tag;
- the fourth video total number statistics submodule is configured to determine the number of video data in the second video set corresponding to each original tag to obtain the fourth total number of videos corresponding to each original tag;
- the preference calculation submodule is set To calculate the user's preference for each original tag based on the total number of fourth videos corresponding to each original tag and the total number of third videos corresponding to each original tag, where the preference It is positively correlated with the total number of fourth videos, and the preference degree is negatively correlated with the
- the positive operation behavior data includes data generated by at least one operation of playing a complete video, liking the video, sharing the video, following a user, and commenting on the video.
- the probability calculation module 403 includes: a system parameter acquisition sub-module configured to acquire a first system parameter and a second system parameter; a first product calculation sub-module configured to calculate the first system parameter and each The product of the saliency value of the original tag is used to obtain the first product corresponding to each original tag; the second product calculation sub-module is configured to calculate the second system parameter and the user's preference for each original tag The second product corresponding to each original tag is obtained; and the sum value calculation sub-module is set to calculate the first product corresponding to each original tag and the second product corresponding to each original tag Sum value; a probability calculation sub-module configured to calculate the probability of each original tag used for negative feedback based on the sum value.
- the candidate tag determination module 404 includes: a sorting sub-module configured to sort the multiple original tags in descending order according to the probability that each original tag is used for negative feedback; the candidate tag determination sub-module is configured to In order to use the top N original tags as candidate tags, where N ⁇ 1.
- a candidate tag pushing module configured to push the candidate tag to the client in the case of detecting the negative feedback operation of the user on the video data, and the client is used to The video playback interface displays the candidate tag;
- the target tag determination module is configured to determine the candidate tag from the candidate tag according to the negative feedback operation in the case of receiving a negative feedback operation of the user on the candidate tag Describe the user’s target tag.
- the label determination device for negative feedback provided by the embodiment of the present application can execute the label determination method for negative feedback provided by any embodiment of the present application, and is equipped with functional modules corresponding to the execution method.
- the video recommendation device in this embodiment of the present application may include the following modules: a user determination module 501 configured to determine the video to be recommended The user; the video data acquisition module 502 is set to acquire a plurality of video data; the user tag acquisition module 503 is set to acquire the target tag of the user; the target video data determination module 504 is set to obtain the target tag from the multiple The target video data is determined from the video data; the video push module 505 is configured to push the target video data to the client; wherein, the target tag is the tag selected from the candidate tags when the user performs a negative feedback operation
- the candidate label is determined by the label determination device for negative feedback described in the fourth embodiment.
- the target video data determining module 504 includes: a first video data determining submodule, configured to determine the first video data having the target tag from the plurality of video data; the target video data determining submodule The module is configured to remove the first video data from a plurality of video data to obtain second video data, and use the second video data as target video data.
- the video recommendation device provided by the embodiment of the present application can execute the video recommendation method provided by any embodiment of the present application, and has functional modules corresponding to the execution method.
- the device may include: a processor 70, a memory 71, a display screen 72 with a touch function, an input device 73, an output device 74, and a communication device 75.
- the number of processors 70 in the device may be one or more, and one processor 70 is taken as an example in FIG. 7.
- the number of memories 71 in the device may be one or more, and one memory 71 is taken as an example in FIG. 7.
- the processor 70, the memory 71, the display screen 72, the input device 73, the output device 74, and the communication device 75 of the device may be connected by a bus or in other ways. In FIG. 7, the connection by a bus is taken as an example.
- the memory 71 can be used to store software programs, computer-executable programs, and modules, such as the program instructions/modules corresponding to the label determination device for negative feedback described in any embodiment of the present application (for example, The original label acquisition module 401, the original label parameter acquisition module 402, the probability calculation module 403, and the candidate label determination module 404 in the label determination device for negative feedback described above, and/or the video recommendation as described in any embodiment of the present application
- the program instructions/modules corresponding to the device for example, the user determination module 501, the video data acquisition module 502, the user tag determination module 503, and the target video data determination module 504 in the above-mentioned video recommendation device).
- the memory 71 may mainly include a storage program area and a storage data area.
- the storage program area may store an operating device and an application program required for at least one function; the storage data area may store data created according to the use of the device, etc.
- the memory 71 may include a high-speed random access memory, and may also include a non-volatile memory, such as at least one magnetic disk storage device, a flash memory device, or other non-volatile solid-state storage devices.
- the memory 71 may include a memory remotely provided with respect to the processor 70, and these remote memories may be connected to the device through a network. Examples of the aforementioned networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.
- the display screen 72 is a display screen 72 with a touch function, and the display screen 72 may be a capacitive screen, an electromagnetic screen or an infrared screen.
- the display screen 72 is set to display data according to instructions of the processor 70, and is also set to receive touch operations on the display screen 72 and send corresponding signals to the processor 70 or other devices.
- the display screen 72 is an infrared screen, it also includes an infrared touch frame.
- the infrared touch frame is arranged around the display screen 72 and can also be set to receive infrared signals and send the infrared signals to the processor 70 or other devices.
- the communication device 75 is configured to establish a communication connection with other devices, and may be a wired communication device and/or a wireless communication device.
- the input device 73 can be configured to receive input digital or character information, and to generate key signal input related to user settings and function control of the device, and can also be configured as a camera for acquiring images and a pickup device for acquiring audio data.
- the output device 74 may include audio equipment such as speakers. The specific composition of the input device 73 and the output device 74 can be set according to actual conditions.
- the processor 70 executes various functional applications and data processing of the device by running software programs, instructions, and modules stored in the memory 71, that is, implements the above-mentioned label determination method and/or video recommendation method for negative feedback.
- the processor 70 when the processor 70 executes one or more programs stored in the memory 71, it implements the label determination method for negative feedback and/or the video recommendation method provided in the embodiment of the present application.
- the embodiment of the present application also provides a computer-readable storage medium.
- the instructions in the storage medium are executed by the processor of the device, the device can execute the label determination method for negative feedback and the method described in the above method embodiment. / Or video recommendation method.
- the description is relatively simple, and for related parts, please refer to the part of the description of the method embodiments.
- the technical solution of this application can be embodied in the form of a software product in essence or a part that contributes to related technologies.
- the computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (Read -Only Memory, ROM), Random Access Memory (RAM), flash memory (FLASH), hard disk or optical disk, etc., including multiple instructions to enable a computer device (which can be a robot, a personal computer, a server, Or a network device, etc.) execute the label determination method for negative feedback described in any embodiment of the present application.
- the multiple units and modules included are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be realized;
- the names of multiple functional units are only for the convenience of distinguishing each other, and are not used to limit the protection scope of this application.
- multiple parts of this application can be implemented by hardware, software, firmware, or a combination thereof.
- multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution device.
- a logic gate circuit for implementing logic functions on data signals Discrete logic circuits, application specific integrated circuits with suitable combinational logic gate circuits, Programmable Gate Array (PGA), Field-Programmable Gate Array (FPGA), etc.
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Multimedia (AREA)
- Data Mining & Analysis (AREA)
- Databases & Information Systems (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Library & Information Science (AREA)
- Computational Linguistics (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
- Two-Way Televisions, Distribution Of Moving Picture Or The Like (AREA)
Abstract
一种用于负反馈的标签确定方法、视频推荐方法、装置、设备和存储介质，用于负反馈的标签确定方法包括：获取视频数据的多个原始标签（S110，S210）；获取每个原始标签的显著值以及获取用户对于每个原始标签的偏好度（S120），其中，每个原始标签的显著值表示具有所述每个原始标签的视频数据在所有视频数据中所占的比例；根据每个原始标签的显著值和所述用户对于所述每个原始标签的偏好度计算所述每个原始标签用于负反馈的概率（S130，S270）；根据每个原始标签用于负反馈的概率从多个所述原始标签中确定出候选标签（S140）。
Description
本申请要求在2019年06月26日提交中国专利局、申请号为201910563725.4的中国专利申请的优先权，该申请的全部内容通过引用结合在本申请中。
本申请实施例涉及视频推荐技术领域，例如涉及一种用于负反馈的标签确定方法、视频推荐方法、用于负反馈的标签确定装置、视频推荐装置、设备和存储介质。
随着科学技术的迅猛发展与进步，互联网成为了人们生活中不可或缺的组成部分，通过互联网人们可以在视频播放平台上观看视频，而视频播放平台为了吸引用户，通常向用户推荐用户感兴趣的视频。
视频推荐算法主要是基于用户的反馈向用户推荐视频，在反馈中，可以向用户提供针对视频的负反馈，从而收集用户对视频不感兴趣的原因以优化推荐算法。相关技术中的一种负反馈方式如图1a和图1b所示，该方式通过设置表达不喜欢或不感兴趣的负反馈按钮，用户通过该负反馈按钮对视频进行负反馈，另一种负反馈方式如图1c所示，负反馈被唤醒后，展示多个包括不喜欢作者、视频无聊、内容重复等负反馈选项以供用户作为负反馈的选择。
如上所述的负反馈方式，在视频内容和用户意图复杂化的情况下，无法从负反馈中确定用户由于视频中哪方面的内容造成用户反感，即对视频的负反馈方式无法结合视频内容和用户意图，负反馈指向性差，导致基于负反馈优化推荐算法后，推荐的视频中仍然存在较大比例的视频使得用户反感。
发明内容
本申请实施例提供一种用于负反馈的标签确定方法、视频推荐方法、用于 负反馈的标签确定装置、视频推荐装置、设备和存储介质，以解决相关技术中对视频的负反馈方式无法结合视频内容和用户意图，负反馈指向性差，导致基于负反馈优化推荐算法后，推荐的视频中仍然存在较大比例的视频使得用户反感的问题。
本申请实施例提供了一种用于负反馈的标签确定方法，包括：
获取视频数据的多个原始标签；
获取每个原始标签的显著值，以及获取用户对于每个原始标签的偏好度，其中，每个原始标签的显著值表示具有所述每个原始标签的视频数据在所有视频数据中所占的比例；
根据每个原始标签的显著值和所述用户对于所述每个原始标签的偏好度计算所述每个原始标签用于负反馈的概率；
根据每个原始标签用于负反馈的概率从所述多个原始标签中确定出候选标签。
本申请实施例提供了一种视频推荐方法，包括：
确定待推荐视频的用户；
获取多个视频数据；
确定所述用户的目标标签；
基于所述目标标签从所述多个视频数据中确定出目标视频数据；
将所述目标视频数据推送至客户端；
其中，所述目标标签为所述用户执行负反馈操作时从候选标签中选择出的标签，所述候选标签通过本申请实施例所述的用于负反馈的标签确定方法确定。
本申请实施例提供了一种用于负反馈的标签确定装置，包括：
原始标签获取模块，设置为获取视频数据的多个原始标签；
原始标签参数获取模块，设置为获取每个原始标签的显著值，以及获取用户对于每个原始标签的偏好度，其中，每个原始标签的显著值表示具有所述每个原始标签的视频数据在所有视频数据中所占的比例；
概率计算模块，设置为根据每个原始标签的显著值和所述用户对于所述每个原始标签的偏好度计算所述每个原始标签用于负反馈的概率；
候选标签确定模块，设置为根据每个原始标签用于负反馈的概率从所述多个原始标签中确定出候选标签。
本申请实施例提供了一种视频推荐装置，包括：
用户确定模块，设置为确定待推荐视频的用户；
视频数据获取模块，设置为获取多个视频数据；
用户标签确定模块，设置为确定所述用户的目标标签；
目标视频数据确定模块，设置为基于所述目标标签从所述多个视频数据中确定出目标视频数据；
视频推送模块，设置为将所述目标视频数据推送至客户端；
其中，所述目标标签为所述用户执行负反馈操作时从候选标签中选择出的标签，所述候选标签通过本申请实施例所述的用于负反馈的标签确定装置确定。
本申请实施例提供了一种设备，所述设备包括：
一个或多个处理器；
存储装置，设置为存储一个或多个程序，
当所述一个或多个程序被所述一个或多个处理器执行，使得所述一个或多个处理器实现本申请任一实施例所述的用于负反馈的标签确定方法和/或视频推荐方法。
本申请实施例提供了一种计算机可读存储介质，其上存储有计算机程序， 该程序被处理器执行时实现本申请任一实施例所述的用于负反馈的标签确定方法和/或视频推荐方法。
图1a是相关技术中的对视频进行负反馈的用户界面的示意图一；
图1b是相关技术中的对视频进行负反馈的用户界面的示意图二；
图1c是相关技术中的对视频进行负反馈的用户界面的示意图三；
图2是本申请实施例一提供的一种用于负反馈的标签确定方法的流程图；
图3A是本申请实施例二提供的一种用于负反馈的标签确定方法的流程图；
图3B是本申请实施例二中视频的候选标签的生成机制的示意图；
图3C是本申请实施例二提供的对视频进行负反馈的用户界面的示意图；
图4是本申请实施例三提供的一种视频推荐方法的流程图；
图5是本申请实施例四提供的一种用于负反馈的标签确定装置的结构框图；
图6是本申请实施例五提供的一种视频推荐装置的结构框图；
图7是本申请实施例六提供的一种设备的结构框图。
下面结合附图和实施例对本申请进行说明。此处所描述的实施例仅仅用于解释本申请，而非对本申请的限定。另外，为了便于描述，附图中仅示出了与本申请相关的部分而非全部结构。
实施例一
图2为本申请实施例一提供的一种用于负反馈的标签确定方法的流程图，本申请实施例可适用于基于标签对视频数据进行负反馈的情况，该方法可以由用于负反馈的标签确定装置来执行，该装置可以通过软件和/或硬件的方式来实现，并集成在执行本方法的设备中，如图2所示，该方法可以包括如下步骤：
S110、获取视频数据的多个原始标签。
原始标签可以是视频数据的关键字、视频的分类信息等，在实际应用中， 可以在离线时预先基于众包的人工标注、基于视觉的自动标注以及基于评论的自动标注等方式为一个视频数据标注标签，在视频数据被播放时，可以获取离线时已为视频数据标注的多个标签作为原始标签。
在本申请的可选实施例中，可以在检测到用户在视频播放平台上播放视频数据时，获取用户播放的视频数据的原始标签。视频播放平台可以是多种短视频播放平台，播放视频数据可以是用户在短视频播放平台选取视频后，在播放页面播放视频数据。
S120、获取每个原始标签的显著值，以及获取用户对于每个原始标签的偏好度。
本申请实施例中，标签的显著值表达了标签对视频内容的区分度，显著值越高，标签对视频内容的区分度越低，显著值越低，标签对视频内容的区分度越高。可选地，显著值可以是标签在所有视频数据中的覆盖率，例如，对于一个标签“学生”，在整个视频播放平台上的所有视频数据中，80％以上的视频数据都有标签“学生”，标签“学生”的覆盖率为80％，说明标签“学生”没有显著性，标签“学生”对视频数据的区分度低，而对于标签“北京男性小学生”，在整个视频播放平台上的所有视频数据中，有10％的视频数据具有标签“北京男性小学生”，标签“北京男性小学生”的覆盖率为10％，说明标签“北京男性小学生”具有显著性，标签“北京男性小学生”对视频数据的区分度高。
标签的偏好度表达了用户对具有该标签的视频数据的感兴趣程度，对于具有一标签的视频数据，如果用户完整地播放了该视频数据，或者是在播放完视频数据后对视频数据执行了关注、分享、评论等正向的操作行为，说明用户对该视频数据感兴趣，在具有该标签的所有视频数据中，具有正向的操作行为的视频数据的比例越高，说明用户对该标签的偏好度越高，用户对具有该标签的视频数据更感兴趣。
在本申请实施例中，可以在离线时计算每个标签的显著值，并且针对每个用户，可以根据用户的历史行为数据计算该用户对每个标签的偏好度。
S130、根据每个原始标签的显著值和所述用户对每个原始标签的偏好度计算所述每个原始标签用于负反馈的概率。
对于当前播放的视频数据的每个原始标签，原始标签用于负反馈的概率与 显著值和偏好度负相关，即显著值和偏好度越小，概率越高。在本申请的可选实施例中，可以计算原始标签的显著值和偏好度的和值，然后计算1与该和值的差值作为原始标签用于负反馈的概率。由于显著值表达了标签对视频内容的区分度，偏好度表达了用户对视频的感兴趣程度，即用户的观看意图，使用标签作为负反馈的标签时，标签的概率越大，通过标签生成的负反馈信息具有更大的概率反映出用户对具有该标签的视频不感兴趣甚至反感。
S140、根据每个原始标签用于负反馈的概率从所述多个原始标签中确定出候选标签。
在确定视频数据的原始标签用于负反馈的概率后，可以选择概率最大的N个原始标签作为候选标签，以供用户从候选标签中选择出目标标签用于对视频数据的负反馈。
例如，在用户播放视频数据时，从视频数据的原始标签确定出候选标签后，如果检测到用户通过长按视频播放界面、点击负反馈按钮等操作方式唤醒对视频数据进行负反馈，可以展示一负反馈页面，并且在负反馈页面展示N个候选标签，以供用户从N个候选标签中选择出候选标签对视频数据进行负反馈，当用户从N个候选标签中选择候选标签进行负反馈时，用户选择的候选标签作为目标标签，将该目标标签与用户关联。
在向用户推荐视频时，可以先获取该用户关联的目标标签，通过目标标签优化推荐算法，例如，可以从所有视频数据中排除掉具有目标标签的视频数据，从而向用户推荐感兴趣的视频数据，避免向用户推荐包含目标标签的视频数据，造成用户不感兴趣甚至反感。
本申请实施例通过获取视频数据的多个原始标签，并获取每个原始标签的显著值以及获取用户对于每个原始标签的偏好度，采用显著值和偏好度计算每个原始标签用于负反馈的概率后，根据概率从多个原始标签中确定出候选标签。本申请实施例中标签的显著值表达了标签对视频内容的区分度，偏好度表达了用户意图，基于显著值和偏好度计算标签用于负反馈的概率，并根据概率确定用于负反馈的标签，综合考虑了视频内容和用户意图，用户从候选标签中选择出目标标签进行负反馈的反馈指向性高，通过用户选择的目标标签优化推荐算法后，能够降低向用户推荐的视频中令用户反感的视频的比例，提高了视频推 荐的准确性。
实施例二
图3A为本申请实施例二提供的一种用于负反馈的标签确定方法的流程图，本申请实施例在实施例一的基础上进行说明，如图3A所示，该方法可以包括如下步骤：
S210、获取所述视频数据的多个原始标签。
S220、统计所有视频数据的数量，得到第一视频总数。
在本申请实施例中，可以获取整个视频播放平台上所有视频数据的视频集合V＝{V
1，V
2，…，V
n}，并且该视频集合V所拥有的标签集合为S
T＝{T
1，T
2，…，T
k}，其中，标签集合S
T为获取视频集合V中的所有视频数据V
i的标签后进行去重后所生成的标签集合。对于视频集合V，可以统计视频集合V中所包含的视频数据的数量作为第一视频总数。
S230、统计具有每个原始标签的视频数据的数量，得到第二视频总数。
可以通过标签和视频数据之间的关联关系，在视频集合V中查找具有一个标签的视频数据，然后统计具有该标签的视频数据的数量作为第二视频总数。
S240、计算所述第二视频总数和所述第一视频总数之间的比值，将所述比值作为所述每个原始标签的显著值。
在本申请实施例中，一个标签的显著值可以为标签的覆盖率，即在视频集合V中，具有该标签的视频数据的比例。因此，标签的显著值可以通过具有该标签的视频数据的数量与所有视频数据的数量的比值来表示，即将计算第二视频总数和第一视频总数得到的比值作为标签的显著值，显著值越大，标签的显著性越低，显著值越小，标签的显著性越高，对于视频标签集合S
T中的每个标签T
i，可以通过以上方式实时计算每个标签T
i的显著值。
S250、获取用户数据，所述用户数据用于表示所述用户在播放视频数据的情况下触发的操作。
在本申请实施例中，用户数据包括视频播放列表和操作行为数据，视频播放列表可以是用户播放过的视频数据的列表，用户在播放视频数据时的操作行为可以是用户在播放视频数据后立即关闭视频数据、播放完视频数据后关闭视 频数据、对视频数据进行评论、关注视频数据、分享视频数据等操作。
在本申请的可选实施例中，可以确定用户请求播放的视频数据，将视频数据记录在视频播放列表中，并记录用户对视频数据触发的操作以生成的操作行为数据。
在实际应用中，可以为每个用户维护一个观看历史列表，该观看历史列表即为用户的视频播放列表V
u＝{V
u,1，V
u,2，…,V
u,m}，视频播放列表V
u中包含了用户播放过的视频数据，例如，用户请求播放视频数据时，将该视频数据记录在视频播放列表V
u中。另外，在用户播放视频数据过程中，可以记录用户的操作生成播放日志作为操作行为数据，该播放日志记录了用户播放视频开始时间、用户播放视频结束时间和在播放过程中执行的操作。
S260、根据所述用户数据确定所述用户对于每个原始标签的偏好度。
在本申请的可选实施例中，可以从视频播放列表中确定出具有每个原始标签的视频数据作为所述每个原始标签对应的第一视频集合，从所述每个原始标签对应的第一视频集合中确定出具有正向的操作行为数据的视频数据作为所述每个原始标签对应的第二视频集合，然后确定所述每个原始标签对应的第一视频集合中视频数据的数量得到所述每个原始标签对应的第三视频总数，以及确定所述每个原始标签对应的第二视频集合中视频数据的数量得到所述每个原始标签对应的第四视频总数，最后基于所述每个原始标签对应的第四视频总数和所述每个原始标签对应的第三视频总数计算用户对于所述每个原始标签的偏好度，其中，偏好度与第四视频总数正相关，与第三视频总数负相关。
正向的操作行为可以是表达用户对一视频数据感兴趣的行为，可选地，正向的操作行为可以是播放完整视频、对视频点赞、分享视频、关注用户、对视频评论中的至少一种。
对于一个原始标签T
i，可以从用户播放过的视频播放列表V
u中确定出具有标签T
i的第一视频集合，以及从第一视频集合中确定出具有正向的操作行为的第二视频集合，然后采用第一视频集合中视频数据的数量总数和第二视频集合中视频数据的数量总数计算偏好度。可以通以下公式计算用户u对标签T
i的偏好度：
其中，M为平滑参数，可以取值为30，由以上公式可知，对于一个标签T
i，在用户播放过的视频数据中，具有标签T
i并且有正向的操作行为的视频数量越大，说明用户对具有标签T
i的视频数据越感兴趣。
S270、根据每个原始标签的显著值和所述用户对所述每个原始标签的偏好度计算所述每个原始标签用于负反馈的概率。
在本申请实施例中，可以先获取第一系统参数和第二系统参数，计算第一系统参数和每个原始标签的显著值的乘积得到第一乘积，计算第二系统参数和用户对所述每个原始标签的偏好度的乘积得到第二乘积，计算第一乘积和第二乘积的和值，计算1与该和值的差值作为所述每个原始标签用于负反馈的概率。
对于用户u当前播放的视频数据V
c，在确定视频数据V
c的多个原始标签
后，可以获取离线计算得到的每个原始标签
的显著值
和偏好度
然后通过以下公式计算原始标签用于负反馈的概率：
其中，
为用户u播放的视频数据V
c的原始标签
用于负反馈的概率，α和β为正值的系统参数，系统参数可以通过在线测试后根据用户的反馈进行调整，在上述公式中，原始标签
的显著值
和偏好度
越小，其用于负反馈的概率
越大。
本申请实施例基于显著值和偏好度计算原始标签用于负反馈的概率，并根据概率从原始标签确定出用于负反馈的候选标签，综合考虑了视频内容和用户意图，用户从候选标签中选择出目标标签对视频数据进行负反馈的反馈指向性高，通过用户选择的目标标签优化推荐算法后，能够降低向用户推荐的视频中令用户反感的视频的比例，提高了视频推荐的准确性。
S280、按照每个原始标签用于负反馈的概率的大小对所述多个原始标签进行降序排序。
对于视频数据V
c的原始标签
在确定每个原始标签
用于负反馈的概率后，可以根据概率的大小对原始标签进行排序，可选地，可以按照概率大小对原始标签进行降序排序，以便确定概率最大的多个原始标 签作为候选标签。
S290、将排序在前的N个原始标签作为候选标签，其中N≥1。
本申请实施例中，原始标签按照概率大小进行降序排序，概率越大说明采用该原始标签作为候选标签进行负反馈时，具有更大的概率收集到准确的负反馈信息，由此可以将排序在前的N个原始标签作为候选标签。
如图3B所示为本申请实施例的当前播放的视频数据的候选标签生成机制，候选标签生成机制如下：
用户在播放视频时，获取当前播放视频的m个标签，并获取离线生成的用户画像，该用户画像记录了用户播放过的n个视频，以及n个视频所具有的n个标签的偏好度，另外获取离线时已经生成的当前播放视频的m个标签的显著值，然后从n个标签的偏好度中查找出当前播放视频的m个标签的偏好度，通过偏好度和显著值计算当前播放视频的m个标签中每个标签用于负反馈的概率，然后将概率最大的3个标签作为候选标签，即候选标签1、候选标签2和候选标签3。
如图3C所示，在确定视频数据用于负反馈的3个候选标签后，当检测到用户唤醒对视频数据的负反馈时，将候选标签推送至客户端，客户端显示负反馈页面，在该负反馈页面中展示3个候选标签以供用户在负反馈页面中选择出目标标签对视频数据进行负反馈，例如，用户在负反馈页面中选择一个候选标签对视频数据进行负反馈，将用户选择的候选标签作为用户的目标标签，以便在向该用户推荐视频时，根据该目标标签优化推荐算法，确定出视频数据推荐给用户。
如图3C所示，候选标签的数量N的取值可以为3，即将概率最大的3个原始标签作为候选标签。如果检测到用户通过长按视频播放界面的方式唤醒负反馈，可以弹出负反馈页面，并在负反馈页面中显示标签1、标签2和标签3共3个候选标签，在检测到用户针对3个候选标签的选择操作时，将用户选择的候选标签作为用户的目标标签，如用户从标签1、标签2和标签3中选择标签1，标签1为目标标签，在向用户推荐视频时，可以采用标签1优化推荐算法，避免向用户推荐具有标签1的视频数据。
本申请实施例中，标签的显著值表达了标签对视频内容的区分度，偏好度 表达了用户意图，基于显著值和偏好度计算原始标签用于负反馈的概率，并根据概率确定用于负反馈的候选标签，在检测到用户针对视频数据的负反馈操作时展示候选标签，然后根据用户针对候选标签的操作，从候选标签中确定出目标标签，在向用户推荐视频时，通过目标标签优化对用户进行视频推荐的推荐算法，使得从候选标签中确定出目标标签进行负反馈后，获得的负反馈信息综合了视频内容和用户意图，负反馈的指向性高，通过目标标签优化推荐算法后能够降低向用户推荐的视频数据中令用户反感的视频数据的比例，提高了视频推荐的准确性。
实施例三
图4为本申请实施例三提供的一种视频推荐方法的流程图，本申请实施例可适用于向用户推荐视频的情况，该方法可以由视频推荐装置来执行，该装置可以通过软件和/或硬件的方式来实现，并集成在执行本方法的设备中，如图4所示，该方法可以包括如下步骤：
S310、确定待推荐视频的用户。
在本申请实施例中，可以在检测到预设事件时向用户推荐视频，预设事件可以是检测到用户通过账号登录短视频播放平台、用户在视频预览界面执行翻页、刷新等操作，如果检测到预设事件，触发预设事件的用户为待推荐视频的用户，当然，也可以根据用户的历史行为对用户进行分类，将一分类用户作为待推荐视频的用户，例如，当有新的视频数据时，如果该视频数据比较合适老人观看，可以将年龄段在50-70岁的用户作为待推荐视频的用户，本申请实施例对确定待推荐视频的用户的方式不加以限制。
S320、获取多个视频数据。
可以根据用户的历史行为数据确定用户感兴趣的视频数据，当然，还可以获取一个时间段内首次上传的视频数据，或者随机获取视频数据，本申请实施例对获取多个视频数据的方式不加以限制。
S330、确定所述用户的目标标签。
在本申请实施例中，目标标签可以是用户在对视频数据进行负反馈时，从视频数据用于负反馈的候选标签中选择的标签，即目标标签为用户不感兴趣甚至反感的视频数据的标签。候选标签可以通过以下方式确定：
S10、在用户播放视频数据的情况下，获取所述视频数据的多个原始标签。
S20、获取每个原始标签的显著值，以及获取所述用户对于每个原始标签的偏好度。
S30、根据每个原始标签的显著值和所述用户对每个原始标签的偏好度计算所述每个原始标签用于负反馈的概率。
S40、根据每个原始标签用于负反馈的概率从所述多个原始标签中确定出候选标签。
在确定候选标签后，当检测到用户唤醒对视频数据的负反馈时，将候选标签推送至客户端，以在客户端的负反馈页面展示候选标签，在接收到用户在负反馈页面上针对候选标签的选择操作时，根据用户的选择操作从展示的候选标签中确定出用户对视频数据进行负反馈的目标标签。
确定候选标签和用户的目标标签的步骤可参考实施例一和实施例二，在此不再详述。
S340、基于所述目标标签从所述多个视频数据中确定出目标视频数据。
可以从多个视频数据中确定出具有目标标签的视频数据作为第一视频数据，然后在多个视频数据中去除第一视频数据得到第二视频数据，将第二视频数据作为目标视频数据。
在实际应用中，本领域技术人员还可以通过其他方式根据目标标签从多个视频数据中确定出用户不感兴趣甚至反感的视频数据，以避免向用户推荐用户不感兴趣的视频数据，本申请实施例对根据目标标签从多个视频数据中确定出目标视频数据的方式不加以限制。
S350、将所述目标视频数据推送至客户端。
在确定目标视频数据后，可以将目标视频数据推送至客户端，以在客户端展示目标视频数据，进而向用户推荐目标视频数据，例如，客户端显示推荐列表，该推荐列表包括多个目标视频数据以供用户选择播放。
本申请实施例中，用户的目标标签为负反馈时用户选择的候选标签，该候选标签基于候选标签的显著值和偏好度计算用于负反馈的概率后确定，由于显著值表达了标签对视频内容的区分度，偏好度表达了用户意图，使得用户从候 选标签中选择出目标标签后，通过目标标签进行负反馈获得的负反馈信息综合了视频内容和用户意图，负反馈的指向性高，在向用户推荐视频时，通过目标标签从多个视频数据中确定出目标视频数据，降低了目标视频数据中令用户反感的视频数据的比例，提高了视频推荐的准确性。
实施例四
图5是本申请实施例四提供的一种用于负反馈的标签确定装置的结构框图，如图5所示，本申请实施例的用于负反馈的标签确定装置可以包括如下模块：原始标签获取模块401，设置为获取视频数据的多个原始标签；原始标签参数获取模块402，设置为获取每个原始标签的显著值，以及获取用户对于每个原始标签的偏好度，其中，每个原始标签的显著值表示具有所述每个原始标签的视频数据在所有视频数据中所占的比例；概率计算模块403，设置为根据每个原始标签的显著值和所述用户对于所述每个原始标签的偏好度计算所述每个原始标签用于负反馈的概率；候选标签确定模块404，设置为根据每个原始标签用于负反馈的概率从所述多个原始标签中确定出候选标签。
可选地，所述原始标签参数获取模块402包括：第一视频总数统计子模块，设置为统计所有视频数据的数量，得到第一视频总数；第二视频总数统计子模块，设置为统计具有每个原始标签的视频数据的数量，得到第二视频总数；显著值计算子模块，设置为计算所述第二视频总数和所述第一视频总数之间的比值，将所述比值作为所述每个原始标签的显著值。
可选地，所述原始标签参数获取模块402包括：用户数据获取子模块，设置为获取用户数据，所述用户数据用于表示所述用户在播放所述视频数据的情况下触发的操作；偏好度确定子模块，设置为根据所述用户数据确定所述用户对于每个原始标签的偏好度。
可选地，所述用户数据包括视频播放列表和操作行为数据，所述用户数据获取子模块包括：视频数据确定单元，设置为确定所述用户请求播放的视频数据；视频数据记录单元，设置为将所述视频数据记录在视频播放列表中；操作行为数据记录单元，设置为记录所述用户对所述视频数据触发的操作，以生成操作行为数据。
可选地，所述原始标签参数获取模块402包括：第一视频集合确定子模块， 设置为从所述视频播放列表中确定出具有每个原始标签的视频数据，作为所述每个原始标签对应的第一视频集合；第二视频集合确定子模块，设置为从所述每个原始标签对应的第一视频集合中确定出具有正向的操作行为数据的视频数据，作为所述每个原始标签对应的第二视频集合；第三视频总数统计子模块，设置为确定所述每个原始标签对应的第一视频集合中视频数据的数量，得到所述每个原始标签对应的第三视频总数；第四视频总数统计子模块，设置为确定所述每个原始标签对应的第二视频集合中视频数据的数量，得到所述每个原始标签对应的第四视频总数；偏好度计算子模块，设置为基于所述每个原始标签对应的第四视频总数和所述所述每个原始标签对应的第三视频总数计算所述用户对于所述每个原始标签的偏好度，其中，所述偏好度与所述第四视频总数正相关，所述偏好度与所述第三视频总数负相关。
可选地，所述正向的操作行为数据包括播放完整视频、对视频点赞、分享视频、关注用户、对视频评论中的至少一种操作所生成的数据。
可选地，所述概率计算模块403包括：系统参数获取子模块，设置为获取第一系统参数和第二系统参数；第一乘积计算子模块，设置为计算所述第一系统参数和每个原始标签的显著值的乘积，得到所述每个原始标签对应的第一乘积；第二乘积计算子模块，设置为计算所述第二系统参数和所述用户对应所述每个原始标签的偏好度的乘积，得到所述每个原始标签对应的第二乘积；和值计算子模块，设置为计算所述每个原始标签对应的第一乘积和所述每个原始标签对应的第二乘积的和值；概率计算子模块，设置为基于所述和值计算所述每个原始标签用于负反馈的概率。
可选地，所述候选标签确定模块404包括：排序子模块，设置为按照每个原始标签用于负反馈的概率的大小对所述多个原始标签进行降序排序；候选标签确定子模块，设置为将排序在前的N个原始标签作为候选标签，其中N≥1。
可选地，还包括：候选标签推送模块，设置为在检测到所述用户针对所述视频数据的负反馈操作的情况下，将所述候选标签推送至客户端，所述客户端用于在视频播放界面展示所述候选标签；目标标签确定模块，设置为在接收到所述用户针对所述候选标签的负反馈操作的情况下，根据所述负反馈操作从所述候选标签中确定出所述用户的目标标签。
本申请实施例所提供的用于负反馈的标签确定装置可执行本申请任意实施例所提供的用于负反馈的标签确定方法，具备执行方法相应的功能模块。
实施例五
图6是本申请实施例五提供的一种视频推荐装置的结构框图，如图6所示，本申请实施例的视频推荐装置可以包括如下模块：用户确定模块501，设置为确定待推荐视频的用户；视频数据获取模块502，设置为获取多个视频数据；用户标签获取模块503，设置为获取所述用户的目标标签；目标视频数据确定模块504，设置为基于所述目标标签从所述多个视频数据中确定出目标视频数据；视频推送模块505，设置为将所述目标视频数据推送至客户端；其中，所述目标标签为所述用户执行负反馈操作时从候选标签中选择的标签，所述候选标签通过实施例四所述的用于负反馈的标签确定装置确定。
可选地，所述目标视频数据确定模块504包括：第一视频数据确定子模块，设置为从所述多个视频数据中确定出具有所述目标标签的第一视频数据；目标视频数据确定子模块，设置为在多个视频数据中去除所述第一视频数据得到第二视频数据，并将所述第二视频数据作为目标视频数据。
本申请实施例所提供的视频推荐装置可执行本申请任意实施例所提供的视频推荐方法，具备执行方法相应的功能模块。
实施例六
参照图7，示出了本申请一个示例中的一种设备的结构示意图。如图7所示，该设备可以包括：处理器70、存储器71、具有触摸功能的显示屏72、输入装置73、输出装置74以及通信装置75。该设备中处理器70的数量可以是一个或者多个，图7中以一个处理器70为例。该设备中存储器71的数量可以是一个或者多个，图7中以一个存储器71为例。该设备的处理器70、存储器71、显示屏72、输入装置73、输出装置74以及通信装置75可以通过总线或者其他方式连接，图7中以通过总线连接为例。
存储器71作为一种计算机可读存储介质，可用于存储软件程序、计算机可执行程序以及模块，如本申请任意实施例所述的用于负反馈的标签确定装置对应的程序指令/模块(例如，上述用于负反馈的标签确定装置中的原始标签获取模块401、原始标签参数获取模块402、概率计算模块403和候选标签确定模块 404)，和/或如本申请任意实施例所述的视频推荐装置对应的程序指令/模块(例如，上述视频推荐装置中的用户确定模块501、视频数据获取模块502、用户标签确定模块503和目标视频数据确定模块504)。存储器71可主要包括存储程序区和存储数据区，其中，存储程序区可存储操作装置、至少一个功能所需的应用程序；存储数据区可存储根据设备的使用所创建的数据等。此外，存储器71可以包括高速随机存取存储器，还可以包括非易失性存储器，例如至少一个磁盘存储器件、闪存器件或其他非易失性固态存储器件。在一些实例中，存储器71可包括相对于处理器70远程设置的存储器，这些远程存储器可以通过网络连接至设备。上述网络的实例包括但不限于互联网、企业内部网、局域网、移动通信网及其组合。
显示屏72为具有触摸功能的显示屏72，显示屏72可以是电容屏、电磁屏或者红外屏。一般而言，显示屏72设置为根据处理器70的指示显示数据，还设置为接收作用于显示屏72的触摸操作，并将相应的信号发送至处理器70或其他装置。可选的，当显示屏72为红外屏时，还包括红外触摸框，该红外触摸框设置在显示屏72的四周，还可以设置为接收红外信号，并将该红外信号发送至处理器70或者其他设备。
通信装置75，设置为与其他设备建立通信连接，可以是有线通信装置和/或无线通信装置。
输入装置73可设置为接收输入的数字或者字符信息，以及产生与设备的用户设置以及功能控制有关的键信号输入，还可以设置为获取图像的摄像头以及获取音频数据的拾音设备。输出装置74可以包括扬声器等音频设备。输入装置73和输出装置74的具体组成可以根据实际情况设定。
处理器70通过运行存储在存储器71中的软件程序、指令以及模块，从而执行设备的多种功能应用以及数据处理，即实现上述用于负反馈的标签确定方法和/或视频推荐方法。
实施例中，处理器70执行存储器71中存储的一个或多个程序时，实现本申请实施例提供的用于负反馈的标签确定方法和/或视频推荐方法。
本申请实施例还提供一种计算机可读存储介质，所述存储介质中的指令由设备的处理器执行时，使得设备能够执行如上述方法实施例所述的用于负反馈 的标签确定方法和/或视频推荐方法。
对于装置、设备、存储介质实施例而言，由于装置、设备、存储介质实施例与方法实施例基本相似，所以描述的比较简单，相关之处参见方法实施例的部分说明即可。
通过以上关于实施方式的描述，所属领域的技术人员可以清楚地了解到，本申请可借助软件及必需的通用硬件来实现，当然也可以通过硬件实现。本申请的技术方案本质上或者说对相关技术做出贡献的部分可以以软件产品的形式体现出来，该计算机软件产品可以存储在计算机可读存储介质中，如计算机的软盘、只读存储器(Read-Only Memory，ROM)、随机存取存储器(Random Access Memory，RAM)、闪存(FLASH)、硬盘或光盘等，包括多个指令用以使得一台计算机设备(可以是机器人，个人计算机，服务器，或者网络设备等)执行本申请任意实施例所述的用于负反馈的标签确定方法。
上述用于负反馈的标签确定装置和视频推荐装置中，所包括的多个单元和模块只是按照功能逻辑进行划分的，但并不局限于上述的划分，只要能够实现相应的功能即可；另外，多个功能单元的名称也只是为了便于相互区分，并不用于限制本申请的保护范围。
应当理解，本申请的多个部分可以用硬件、软件、固件或它们的组合来实现。在上述实施方式中，多个步骤或方法可以用存储在存储器中且由合适的指令执行装置执行的软件或固件来实现。例如，如果用硬件来实现，和在另一实施方式中一样，可用本领域公知的下列技术中的任一项或他们的组合来实现：具有用于对数据信号实现逻辑功能的逻辑门电路的离散逻辑电路，具有合适的组合逻辑门电路的专用集成电路，可编程门阵列(Programmable Gate Array，PGA)，现场可编程门阵列(Field-Programmable Gate Array，FPGA)等。
在本说明书的描述中，参考术语“一个实施例”、“一些实施例”、“示例”、“具体示例”或“一些示例”等的描述意指结合该实施例或示例描述的具体特征、结构、材料或者特点包含于本申请的至少一个实施例或示例中。在本说明书中，对上述术语的示意性表述不一定指的是相同的实施例或示例。而且，描述的具体特征、结构、材料或者特点可以在任何的一个或多个实施例或示例中以合适的方式结合。
Claims (15)
- 一种用于负反馈的标签确定方法，包括：获取视频数据的多个原始标签；获取每个原始标签的显著值，以及获取用户对于每个原始标签的偏好度，其中，每个原始标签的显著值表示具有所述每个原始标签的视频数据在所有视频数据中所占的比例；根据每个原始标签的显著值和所述用户对于所述每个原始标签的偏好度计算所述每个原始标签用于负反馈的概率；根据每个原始标签用于负反馈的概率从所述多个原始标签中确定出候选标签。
- 如权利要求1所述的方法，其中，所述获取每个原始标签的显著值，包括：统计所有视频数据的数量，得到第一视频总数；统计具有每个原始标签的视频数据的数量，得到第二视频总数；计算所述第二视频总数和所述第一视频总数之间的比值，将所述比值作为所述每个原始标签的显著值。
- 如权利要求1所述的方法，其中，所述获取用户对于每个原始标签的偏好度，包括：获取用户数据，所述用户数据用于表示所述用户在播放所述视频数据的情况下触发的操作；根据所述用户数据确定所述用户对于每个原始标签的偏好度。
- 如权利要求3所述的方法，其中，所述用户数据包括视频播放列表和操作行为数据，所述获取用户数据，包括：确定所述用户请求播放的视频数据；将所述视频数据记录在视频播放列表中；记录所述用户对所述视频数据触发的操作，以生成操作行为数据。
- 如权利要求4所述的方法，其中，所述根据所述用户数据确定所述用户对于每个原始标签的偏好度，包括：从所述视频播放列表中确定出具有每个原始标签的视频数据，作为所述每 个原始标签对应的第一视频集合；从所述每个原始标签对应的第一视频集合中确定出具有正向的所述操作行为数据的视频数据，作为所述每个原始标签对应第二视频集合；确定所述每个原始标签对应的第一视频集合中视频数据的数量，得到所述每个原始标签对应的第三视频总数；确定所述每个原始标签对应的第二视频集合中视频数据的数量，得到所述每个原始标签对应的第四视频总数；基于所述每个原始标签对应的第四视频总数和所述每个原始标签对应的第三视频总数计算所述用户对于所述每个原始标签的偏好度，其中，所述偏好度与所述第四视频总数正相关，所述偏好度与所述第三视频总数负相关。
- 如权利要求5所述的方法，其中，所述正向的操作行为数据包括播放完整视频、对视频点赞、分享视频、关注用户、对视频评论中的至少一种操作所生成的数据。
- 如权利要求1所述的方法，其中，所述根据每个原始标签的显著值和所述用户对于每个原始标签的偏好度计算所述每个原始标签用于负反馈的概率，包括：获取第一系统参数和第二系统参数；计算所述第一系统参数和每个原始标签的显著值的乘积，得到所述每个原始标签对应的第一乘积；计算所述第二系统参数和所述用户对应所述每个原始标签的偏好度的乘积，得到所述每个原始标签对应的第二乘积；计算所述每个原始标签对应的第一乘积和所述每个原始标签对应的第二乘积的和值；基于所述和值计算所述每个原始标签用于负反馈的概率。
- 如权利要求1-7中任一项所述的方法，其中，所述根据每个原始标签用于负反馈的概率从所述多个原始标签中确定出候选标签，包括：按照每个原始标签用于负反馈的概率的大小对所述多个原始标签进行降序排序；将排序在前的N个原始标签作为候选标签，其中N≥1。
- 如权利要求1-7中任一项所述的方法，还包括：在检测到所述用户针对所述视频数据的负反馈操作的情况下，将所述候选标签推送至客户端，所述客户端用于展示所述候选标签；在接收到所述用户针对所述候选标签的负反馈操作的情况下，根据所述负反馈操作从所述候选标签中确定出所述用户的目标标签。
- 一种视频推荐方法，包括：确定待推荐视频的用户；获取多个视频数据；确定所述用户的目标标签；基于所述目标标签从所述多个视频数据中确定出目标视频数据；将所述目标视频数据推送至客户端；其中，所述目标标签为所述用户执行负反馈操作时从候选标签中选择出的标签，所述候选标签通过如权利要求1-8中任一项所述的用于负反馈的标签确定方法确定。
- 如权利要求10所述的方法，其中，所述基于所述目标标签从所述多个视频数据中确定出目标视频数据，包括：从所述多个视频数据中确定出具有所述目标标签的第一视频数据；在多个视频数据中去除所述第一视频数据得到第二视频数据，并将所述第二视频数据作为目标视频数据。
- 一种用于负反馈的标签确定装置，包括：原始标签获取模块，设置为获取视频数据的多个原始标签；原始标签参数获取模块，设置为获取每个原始标签的显著值，以及获取用户对于每个原始标签的偏好度，其中，每个原始标签的显著值表示具有所述每个原始标签的视频数据在所有视频数据中所占的比例；概率计算模块，设置为根据每个原始标签的显著值和所述用户对所述每个原始标签的偏好度计算所述每个原始标签用于负反馈的概率；候选标签确定模块，设置为根据每个原始标签用于负反馈概率从所述多个 原始标签中确定出候选标签。
- 一种视频推荐装置，包括：用户确定模块，设置为确定待推荐视频的用户；视频数据获取模块，设置为获取多个视频数据；用户标签确定模块，设置为确定所述用户的目标标签；目标视频数据确定模块，设置为基于所述目标标签从所述多个视频数据中确定出目标视频数据；视频推送模块，设置为将所述目标视频数据推送至客户端；其中，所述目标标签为所述用户执行负反馈操作时从候选标签中选择出的标签，所述候选标签通过如权利要求12所述的用于负反馈的标签确定装置确定。
- 一种设备，包括：至少一个处理器；存储装置，设置为存储至少一个程序，当所述至少一个程序被所述至少一个处理器执行，使得所述至少一个处理器实现以下至少之一：如权利要求1-9中任一项所述的用于负反馈的标签确定方法；如权利要求10-11中任一项所述的视频推荐方法。
- 一种计算机可读存储介质，存储有计算机程序，该程序被处理器执行时实现以下至少之一：如权利要求1-9中任一项所述的用于负反馈的标签确定方法；如权利要求10-11中任一项所述的视频推荐方法。
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910563725.4A CN110287372B (zh) | 2019-06-26 | 2019-06-26 | 用于负反馈的标签确定方法、视频推荐方法及其装置 |
| CN201910563725.4 | 2019-06-26 |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| WO2020259572A1 true WO2020259572A1 (zh) | 2020-12-30 |
Family
ID=68007539
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| PCT/CN2020/098065 WO2020259572A1 (zh) | 2019-06-26 | 2020-06-24 | 用于负反馈的标签确定方法、视频推荐方法、装置、设备和存储介质 |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN110287372B (zh) |
| WO (1) | WO2020259572A1 (zh) |
Families Citing this family (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110287372B (zh) * | 2019-06-26 | 2021-06-01 | 广州市百果园信息技术有限公司 | 用于负反馈的标签确定方法、视频推荐方法及其装置 |
| CN112711703B (zh) * | 2019-10-25 | 2024-03-26 | 北京达佳互联信息技术有限公司 | 用户标签获取方法、装置、服务器及存储介质 |
| CN110941740B (zh) * | 2019-11-08 | 2023-07-14 | 深圳市雅阅科技有限公司 | 视频推荐方法及计算机可读存储介质 |
| CN111209432A (zh) * | 2020-01-02 | 2020-05-29 | 北京字节跳动网络技术有限公司 | 信息获取的方法、装置、电子设备及计算机可读介质 |
| CN111291688B (zh) * | 2020-02-12 | 2023-07-14 | 咪咕文化科技有限公司 | 视频标签的获取方法及装置 |
| CN111522993B (zh) * | 2020-04-08 | 2023-08-15 | 咪咕文化科技有限公司 | 标签有效性管理方法、装置、网络设备及存储介质 |
| CN112188295B (zh) * | 2020-09-29 | 2022-07-05 | 有半岛(北京)信息科技有限公司 | 一种视频推荐方法及装置 |
| CN113537215A (zh) * | 2021-07-19 | 2021-10-22 | 山东福来克思智能科技有限公司 | 一种视频标签标注的方法及设备 |
| CN115618101A (zh) * | 2022-09-21 | 2023-01-17 | 清华大学 | 基于负反馈的流媒体内容推荐方法、装置及电子设备 |
| CN116010650B (zh) * | 2023-03-20 | 2023-09-15 | 南京邮电大学 | 一种基于负反馈的用户反感标签确定方法 |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101141591A (zh) * | 2007-09-29 | 2008-03-12 | 腾讯科技（深圳）有限公司 | 节目列表显示方法及网络视频系统 |
| US20130195422A1 (en) * | 2012-02-01 | 2013-08-01 | Cisco Technology, Inc. | System and method for creating customized on-demand video reports in a network environment |
| CN103729785A (zh) * | 2014-01-26 | 2014-04-16 | 合一信息技术（北京）有限公司 | 一种视频用户性别分类方法及其装置 |
| CN105142028A (zh) * | 2015-07-29 | 2015-12-09 | 华中科技大学 | 面向三网融合的电视节目内容搜索与推荐方法 |
| WO2017128054A1 (zh) * | 2016-01-26 | 2017-08-03 | 白宏伟 | 视频自动更新技术的数据采集方法以及视频系统 |
| CN110287372A (zh) * | 2019-06-26 | 2019-09-27 | 广州市百果园信息技术有限公司 | 用于负反馈的标签确定方法、视频推荐方法及其装置 |
Family Cites Families (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20130046761A1 (en) * | 2010-01-08 | 2013-02-21 | Telefonaktiebolaget L M Ericsson (Publ) | Method and Apparatus for Social Tagging of Media Files |
| KR20130125047A (ko) * | 2012-05-08 | 2013-11-18 | 삼성전자주식회사 | 디스플레이장치 및 컨텐츠 영상 제공장치 및 이들의 제어방법 |
| CN105045907B (zh) * | 2015-08-10 | 2018-03-09 | 北京工业大学 | 一种用于个性化社会图像推荐的视觉注意‑标签‑用户兴趣树的构建方法 |
| CN108881339B (zh) * | 2017-05-11 | 2022-05-13 | 腾讯科技（深圳）有限公司 | 推送方法、用户标签的生成方法、装置及设备 |
| CN109657138B (zh) * | 2018-12-10 | 2021-02-26 | 深圳大宇无限科技有限公司 | 一种视频推荐方法、装置、电子设备及存储介质 |
-
2019
- 2019-06-26 CN CN201910563725.4A patent/CN110287372B/zh active Active
-
2020
- 2020-06-24 WO PCT/CN2020/098065 patent/WO2020259572A1/zh active Application Filing
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101141591A (zh) * | 2007-09-29 | 2008-03-12 | 腾讯科技（深圳）有限公司 | 节目列表显示方法及网络视频系统 |
| US20130195422A1 (en) * | 2012-02-01 | 2013-08-01 | Cisco Technology, Inc. | System and method for creating customized on-demand video reports in a network environment |
| CN103729785A (zh) * | 2014-01-26 | 2014-04-16 | 合一信息技术（北京）有限公司 | 一种视频用户性别分类方法及其装置 |
| CN105142028A (zh) * | 2015-07-29 | 2015-12-09 | 华中科技大学 | 面向三网融合的电视节目内容搜索与推荐方法 |
| WO2017128054A1 (zh) * | 2016-01-26 | 2017-08-03 | 白宏伟 | 视频自动更新技术的数据采集方法以及视频系统 |
| CN110287372A (zh) * | 2019-06-26 | 2019-09-27 | 广州市百果园信息技术有限公司 | 用于负反馈的标签确定方法、视频推荐方法及其装置 |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110287372B (zh) | 2021-06-01 |
| CN110287372A (zh) | 2019-09-27 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| WO2020259572A1 (zh) | 用于负反馈的标签确定方法、视频推荐方法、装置、设备和存储介质 | |
| US20220256210A1 (en) | Video classification and preview selection | |
| KR102436734B1 (ko) | 비디오 재생 노드 위치 확정 방법, 장치, 전자 장비, 컴퓨터 판독가능 저장 매체 및 컴퓨터 프로그램 | |
| US8995823B2 (en) | Method and system for content relevance score determination | |
| CN103718166B (zh) | 信息处理设备、信息处理方法 | |
| CN1538351B (zh) | 为视频序列生成视频缩略图的方法和计算机 | |
| WO2017092257A1 (zh) | 一种现场直播中的共同收看仿真方法和装置 | |
| US8996983B2 (en) | Device and a method for annotating content | |
| CN112364202B (zh) | 视频推荐方法、装置及电子设备 | |
| WO2021031917A1 (zh) | 视频处理方法、装置、电子设备及可读介质 | |
| US20150052087A1 (en) | Predicting Reactions to Short-Text Posts | |
| WO2019128941A1 (zh) | 新视频推送控制方法、装置及服务器 | |
| WO2022111072A1 (zh) | 图像拍摄方法、装置、电子设备、服务器及存储介质 | |
| WO2018133725A1 (zh) | 媒体文件的输出处理方法、客户端及系统 | |
| CN110225398B (zh) | 多媒体对象播放方法、装置和设备及计算机存储介质 | |
| US10733249B1 (en) | Machine learning system for data selection | |
| US9325754B2 (en) | Information processing device and information processing method | |
| US20220303735A1 (en) | Providing a summary of media content to a communication device | |
| WO2017156994A1 (zh) | 多媒体资源的质量评估方法和装置 | |
| CN109245989A (zh) | 一种基于信息分享的处理方法、装置及计算机可读存储介质 | |
| US11455693B2 (en) | Visual focal point composition for media capture based on a target recipient audience | |
| CN109688217B (zh) | 一种消息推送方法、装置及电子设备 | |
| KR102574278B1 (ko) | 비디오 타임 앵커 | |
| CN111988642B (zh) | 推荐视频的方法、装置、服务器和存储介质 | |
| CN116049490A (zh) | 素材搜索方法、装置和电子设备 |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| 121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 20833135 Country of ref document: EP Kind code of ref document: A1 |
|
| NENP | Non-entry into the national phase |
Ref country code: DE |
|
| 122 | Ep: pct application non-entry in european phase |
Ref document number: 20833135 Country of ref document: EP Kind code of ref document: A1 |