CN108614856B

CN108614856B - Video sequencing calibration method and device

Info

Publication number: CN108614856B
Application number: CN201810236970.XA
Authority: CN
Inventors: 钱士才
Original assignee: Beijing QIYI Century Science and Technology Co Ltd
Current assignee: Beijing QIYI Century Science and Technology Co Ltd
Priority date: 2018-03-21
Filing date: 2018-03-21
Publication date: 2021-01-05
Anticipated expiration: 2038-03-21
Also published as: CN108614856A

Abstract

The embodiment of the invention provides a video sequencing calibration method and device, and relates to the technical field of data processing. Wherein, the method comprises the following steps: determining a play time median and a long click rate of a video pair in a preset time period based on a video pair consisting of a first video and a second video; determining a first calibration parameter of a video pair based on the playing time length and a second calibration parameter based on the long click rate; when an initial recommendation list established for the first video comprises a second video, determining an initial sequencing parameter of the second video; calibrating the initial sequencing parameter of the second video according to the first calibration parameter and the second calibration parameter; the position of the second video in the initial recommendation list is adjusted. The method and the device can calibrate the sorting parameters of the second video in the current recommendation list according to the playing data of the second video in the historical recommendation list, and can improve the click rate and the playing time of the video by utilizing the historical playing data of the playing time and the long click rate for calibration.

Description

Video sequencing calibration method and device

Technical Field

The invention relates to the technical field of data processing, in particular to a video sequencing calibration method and device.

Background

The recommendation system is a service system provided for solving the problem of information overload, and can select resources meeting the interest preference or the demand of a user from a large amount of information for recommendation. With the popularization and rapid development of the internet, recommendation systems have been widely applied to various fields such as video recommendation, news recommendation, and the like.

Generally, when a video is recommended by a video recommendation system, video recall and video ranking need to be performed, wherein when the video recommendation system performs video ranking, influences of various aspects on the video ranking, such as video title search popularity, user interests and the like, are generally considered comprehensively, that is, more characteristic parameters are considered in a video ranking model.

However, although the more feature parameters, the better the model effect, the more features the model often has difficulty in highlighting the most desirable features, and therefore, the video ranking model considering more features is used for video ranking, and the ranking result is difficult to meet the requirements of improving the video click rate and the playing time.

Disclosure of Invention

In view of the foregoing problems, embodiments of the present invention are provided to provide a method and an apparatus for calibrating video ranking, so as to solve the problem that the ranking result of video ranking performed by a video ranking model considering more features is difficult to meet the requirement of improving the video click rate and the playing time.

According to a first aspect of the present invention, there is provided a video ordering calibration method, the method comprising:

determining a play time median and a long click rate of a video pair in a preset time period based on the video pair formed by a first video corresponding to a second video; the second video is a recommended video corresponding to the first video during playing and is played adjacent to the first video within the preset time period;

according to the play time length median value, determining a first calibration parameter of the video pair based on play time length, and according to the long click rate, determining a second calibration parameter of the video pair based on the long click rate;

when the second video is included in the initial recommendation list established for the first video, determining an initial ranking parameter of the second video;

calibrating the initial sequencing parameters of the second video according to the first calibration parameters and the second calibration parameters to obtain calibrated sequencing parameters of the second video;

and adjusting the position of the second video in the initial recommendation list according to the calibrated sorting parameter of the second video.

Optionally, before determining the first calibration parameter of the video pair based on the playing time length according to the playing time length median and determining the second calibration parameter of the video pair based on the long click rate according to the long click rate, the method further includes:

determining the playing times and the long click times of the video pairs in the preset time period;

correspondingly, the determining a first calibration parameter of the video pair based on the playing time length according to the playing time length median value and determining a second calibration parameter of the video pair based on the long click rate according to the long click rate includes:

and determining a first calibration parameter of the video pair based on the playing time length according to the playing time length median and the playing times, and determining a second calibration parameter of the video pair based on the long click rate according to the long click rate and the long click times.

Optionally, before determining the median play time length and the long click rate of the video pair in the preset time period based on the video pair composed of the first video and the second video, the method further includes:

acquiring preset number of sample playing time lengths and sample long click rates;

fitting to obtain a first cumulative distribution model based on the playing duration by taking the sample playing duration as an input parameter;

fitting to obtain a second cumulative distribution model based on the long click rate by taking the sample long click rate as an input parameter;

according to the play time median value, a first calibration parameter of the video pair based on play time is determined through the first cumulative distribution model, and according to the long click rate, a second calibration parameter of the video pair based on the long click rate is determined through the second cumulative distribution model.

acquiring the preset number of sample playing times and the preset number of sample long clicking times;

fitting to obtain a third cumulative distribution model based on the playing times by taking the sample playing times as input parameters;

fitting to obtain a fourth cumulative distribution model based on the long click times by taking the sample long click times as input parameters;

correspondingly, the determining, according to the median playing time length, a first calibration parameter of the video pair based on the playing time length through the first cumulative distribution model, and determining, according to the long click rate, a second calibration parameter of the video pair based on the long click rate through the second cumulative distribution model includes:

according to the playing time length median and the playing times, a first calibration parameter of the video pair based on the playing time length is determined through the first cumulative distribution model and the third cumulative distribution model, and according to the long click rate and the long click times, a second calibration parameter of the video pair based on the long click rate is determined through the second cumulative distribution model and the fourth cumulative distribution model.

Optionally, the determining, based on a video pair formed by a first video and a second video, a median play time length and a long click rate of the video pair in a preset time period includes:

determining a median value of each playing time length of a second video in a preset time period as a median value of the playing time lengths of the video pairs in the preset time period based on a video pair formed by the first video and the second video;

determining the data number of the target playing time length which is greater than the preset playing time length or the ratio of the total time length of the second video to the preset ratio from all the playing time lengths;

and determining the ratio of the data number of the target playing time length to the display number of the second video as the long click rate of the video pair in the preset time period.

According to a second aspect of the present invention, there is provided a video sequencing calibration apparatus, the apparatus comprising:

the first determining module is used for determining a play time median and a long click rate of a video pair in a preset time period based on the video pair formed by a first video corresponding to a second video; the second video is a recommended video corresponding to the first video during playing and is played adjacent to the first video within the preset time period;

the second determining module is used for determining a first calibration parameter of the video pair based on the playing time length according to the median value of the playing time length and determining a second calibration parameter of the video pair based on the long click rate according to the long click rate;

a third determining module, configured to determine an initial ranking parameter of the second video when the second video is included in an initial recommendation list established for the first video;

the calibration module is used for calibrating the initial sequencing parameters of the second video according to the first calibration parameters and the second calibration parameters to obtain the calibrated sequencing parameters of the second video;

and the adjusting module is used for adjusting the position of the second video in the initial recommendation list according to the calibration sorting parameter of the second video.

Optionally, the apparatus further comprises:

the fourth determining module is used for determining the playing times and the long click times of the video pairs in the preset time period;

accordingly, the second determining module comprises:

the first determining submodule is used for determining a first calibration parameter of the video pair based on the playing time length according to the playing time length median and the playing times, and determining a second calibration parameter of the video pair based on the long click rate according to the long click rate and the long click times.

Optionally, the apparatus further comprises:

the first acquisition module is used for acquiring the preset number of sample playing time lengths and sample long click rates;

the first fitting module is used for fitting to obtain a first cumulative distribution model based on the playing time length by taking the sample playing time length as an input parameter;

the second fitting module is used for fitting to obtain a second cumulative distribution model based on the long click rate by taking the sample long click rate as an input parameter;

accordingly, the second determining module comprises:

and the second determining submodule is used for determining a first calibration parameter of the video pair based on the playing time length through the first cumulative distribution model according to the playing time length median value, and determining a second calibration parameter of the video pair based on the long click rate through the second cumulative distribution model according to the long click rate.

Optionally, the apparatus further comprises:

the second acquisition module is used for acquiring the preset number of sample playing times and the preset number of sample long-click times;

the third fitting module is used for fitting to obtain a third cumulative distribution model based on the playing times by taking the sample playing times as an input parameter;

the fourth fitting module is used for fitting to obtain a fourth cumulative distribution model based on the long click times by taking the sample long click times as an input parameter;

correspondingly, the second determining submodule includes:

a determining unit, configured to determine, according to the median of the playing time duration and the playing times, a first calibration parameter of the video pair based on the playing time duration through the first cumulative distribution model and the third cumulative distribution model, and determine, according to the long click rate and the long click times, a second calibration parameter of the video pair based on the long click rate through the second cumulative distribution model and the fourth cumulative distribution model.

Optionally, the first determining module includes:

the third determining submodule is used for determining a median value of playing time lengths of the second video in a preset time period as a median value of the playing time lengths of the video pair in the preset time period based on a video pair formed by the second video corresponding to the first video;

a fourth determining submodule, configured to determine, from the respective play durations, a data number of a target play duration that is greater than a preset play duration or a ratio of the total duration of the second video to the total duration of the second video is greater than a preset ratio;

and the fifth determining submodule is used for determining the ratio of the data number of the target playing time length to the display number of the second video as the long click rate of the video pair in the preset time period.

An embodiment of the present invention further provides a computer-readable storage medium, where a computer program is stored on the computer-readable storage medium, and when the computer program is executed by a processor, the steps of the video sequencing calibration method in the present invention are implemented.

The embodiment of the invention has the following advantages:

based on a video pair composed of a first video and a second video, determining a play time length median and a long click rate of the video pair in a preset time period, wherein the second video is a recommended video corresponding to the first video when playing and is played adjacent to the first video in the preset time period, then determining a first calibration parameter of the video pair based on the play time length according to the play time length median, and determining a second calibration parameter of the video pair based on the long click rate according to the long click rate, when an initial recommendation list established for the first video includes the second video, determining an initial sorting parameter of the second video, calibrating the initial sorting parameter of the second video according to the first calibration parameter and the second calibration parameter to obtain a calibrated sorting parameter of the second video, and then calibrating the sorting parameter according to the second video, the position of the second video in the initial recommendation list is adjusted. In the embodiment of the invention, the sorting parameters of the second video in the current recommendation list of the first video can be calibrated according to the playing data of the second video in the historical recommendation list of the first video, so that the position of the second video in the current recommendation list can be adjusted, the historical playing data of the playing time and the long click rate are utilized for calibration, and the click rate and the playing time of the video can be improved during video recommendation.

Drawings

Fig. 1 is a flowchart of a video sequencing calibration method according to an embodiment of the present invention;

FIG. 2 is a flow chart of another video sequencing calibration method according to an embodiment of the present invention;

fig. 3 is a cumulative distribution curve based on sample playing times according to an embodiment of the present invention;

fig. 4 is a block diagram of a video sequencing calibration apparatus according to an embodiment of the present invention;

fig. 5 is a block diagram of another video sequencing calibration apparatus according to an embodiment of the present invention.

Detailed Description

In order to make the aforementioned objects, features and advantages of the present invention comprehensible, embodiments accompanied with figures are described in further detail below.

Example one

Referring to fig. 1, a flowchart of a video sequencing calibration method is shown, and the method may specifically include the following steps:

step 101: determining a play time median and a long click rate of a video pair in a preset time period based on the video pair formed by the first video corresponding to the second video; the second video is a recommended video corresponding to the first video during playing and is played adjacent to the first video within the preset time period.

The preset time period is a time period which is before the current time and satisfies a preset duration, for example, within 7 days before the current time, within 10 days before the current time, and the like, that is, with respect to the current time, the preset time period is a historical time period, and accordingly, the playing data in the preset time period is also historical playing data. In a video website managed by the data processing device, a first video is any watched video, a second video is any recommended video corresponding to the first video during playing, and the second video is played adjacent to the first video within the preset time period, that is, the second video belongs to any recommended list corresponding to the first video during playing within the preset time period and is played immediately after the first video.

In a video website managed by the data processing device, a large number of registered user objects and a large number of videos generally exist, after each object watches one video, the video terminal can determine the playing time length of the object corresponding to the video, and then the playing time length data can be fed back to the data processing device, so that the data processing device can acquire the playing time length data of each object corresponding to each video in real time. Based on a video pair composed of the first video and the second video, when the second video is played immediately after the first video is played, the playing time length corresponding to the second video may be determined as the playing time length of the video pair. In a preset time period, a plurality of user objects watch the second video immediately after watching the first video, so that in the preset time period, the second video correspondingly generates a plurality of playing time length data, correspondingly, the video pair also corresponds to a plurality of playing time length data in the preset time period, and thus the data processing equipment can acquire each playing time length of the video pair in the preset time period. When video sequencing calibration is required, the data processing device may determine each play time length data of the video pair in a preset time period, and determine a median value of each play time length, that is, a median value of the play time lengths of the video pair in the preset time period.

In addition, the long click may indicate that the user watches the second video for a longer time based on a video pair composed of the first video and the second video, that is, the long click may be considered as one effective click for the video. In practical application, when the total duration of the second video is less than or equal to 150 seconds, the playing duration of the second video reaches more than 80% of the total duration, that is, one long click for the second video is counted, and when the total duration of the second video is greater than 150 seconds, the playing duration of the second video reaches more than 120 seconds, that is, one long click for the second video is counted. In practical applications, the video playing degree defining the long click may be set according to practical situations, and is not limited to the above 80% and 120 seconds, which is not specifically limited in this embodiment of the present invention.

When video sequencing calibration is required, based on a video pair formed by a first video and a second video, the data processing device can determine the long click number and the display number of the video pair in a preset time period, determine the ratio of the long click number to the display number, and obtain the long click rate of the video pair in the preset time period. When the first video and the second video appear as the video pair, the display is regarded as one-time display, and correspondingly, the display number is the number of times of the first video and the second video as the video pair.

Step 102: and determining a first calibration parameter of the video pair based on the playing time length according to the median value of the playing time length, and determining a second calibration parameter of the video pair based on the long click rate according to the long click rate.

The data processing device may use the median of the playing time duration of the video pair as an input parameter, and determine, through a preset first cumulative distribution model, a first calibration parameter of the video pair based on the playing time duration, that is, after the median of the playing time duration is input into the preset first cumulative distribution model, an output value of the first cumulative distribution model is the first calibration parameter of the video pair based on the playing time duration.

In addition, the data processing device may further determine, by using the long click rate of the video pair as an input parameter, a second calibration parameter of the video pair based on the long click rate through a preset second cumulative distribution model, that is, after the long click rate is input to the preset second cumulative distribution model, an output value of the second cumulative distribution model is the second calibration parameter of the video pair based on the long click rate.

Step 103: when the second video is included in the initial recommendation list established for the first video, the initial ranking parameter of the second video is determined.

When a recommendation list needs to be established for a first video, the data processing device may first recall a plurality of recalled videos for the first video, and then score the plurality of recalled videos through a preset ranking model, such as an LR (Logistic Regression) model, a GBDT (Gradient Boosting decision Tree) + LR model, and the like, to obtain an initial ranking parameter of each recalled video, and then rank the plurality of recalled videos according to a sequence of the initial ranking parameters from large to small, so as to obtain an initial recommendation list for the first video. When the second video is included in the initial recommendation list established for the first video, the data processing apparatus may determine an initial ranking parameter of the second video from the initial recommendation list.

Step 104: and calibrating the initial sequencing parameters of the second video according to the first calibration parameters and the second calibration parameters to obtain the calibrated sequencing parameters of the second video.

And the initial sorting parameter of the second video is the sorting score of the second video in the initial recommendation list corresponding to the first video before calibration. When a recommendation list needs to be established for a first video, the data processing device may first recall a plurality of recalled videos for the first video, then score the plurality of recalled videos through a preset ranking model, for example, an LR (Logistic Regression) model, a GBDT (Gradient Boosting decision Tree) + LR model, and the like, a score obtained by scoring each recalled video is an initial ranking parameter of each recalled video, and then rank the plurality of recalled videos according to a sequence of the initial ranking parameters from large to small, so that an initial recommendation list for the first video can be obtained. When the second video is included in the initial recommendation list established for the first video, the data processing apparatus may determine an initial ranking parameter of the second video from the initial recommendation list.

The data processing device may calibrate the initial sorting parameter of the second video by using the first calibration parameter based on the play duration and the second calibration parameter based on the long click rate, so as to obtain the calibrated sorting parameter of the second video.

Step 105: and adjusting the position of the second video in the initial recommendation list according to the calibrated sorting parameter of the second video.

After the calibration, the ranking parameter of the second video changes, but the ranking parameters of other videos in the initial recommendation list do not change, so that the data processing device can adjust the position of the second video in the initial recommendation list according to the calibrated ranking parameter of the second video, and can obtain the recommendation list calibrated for the first video.

The embodiment of the invention has the following advantages:

Example two

Referring to fig. 2, a flow chart of another video sequencing calibration method is shown, and the method may specifically include the following steps:

step 201: and obtaining the preset number of sample playing time lengths and sample long click rates.

The data processing device can obtain a preset number of sample playing time lengths as a fitting input parameter of the first cumulative distribution model based on the playing time lengths, and can obtain a preset number of sample long click rates as a fitting input parameter of the second cumulative distribution model based on the long click rates. In practical application, the sample playing time length and the sample long click rate can be randomly sampled from historical data of a video website by adopting a random sampling method.

For example, the preset number may be 20 ten thousand, and the data processing device may randomly sample the historical data of the video website to obtain 20 ten thousand sample playing time lengths and 20 ten thousand sample long click rates.

Step 202: and fitting to obtain a first cumulative distribution model based on the playing time length by taking the sample playing time length as an input parameter.

The data processing device may obtain a preset number of sample playing durations, draw cumulative distribution discrete data of the sample playing durations by using the sample playing durations as input parameters, then fit the cumulative distribution discrete data to obtain a cumulative distribution curve, and determine a cumulative distribution function corresponding to the cumulative distribution curve, where the cumulative distribution function is a first cumulative distribution model based on the playing durations.

For example, the data processing apparatus may fit a first cumulative distribution model sigmod1 based on the play time length as shown in the following formula (1) with 20 ten thousand sample play time lengths as input parameters.

In the above formula (1), x is the median of the playing time lengths of the second video to be input, and each of the raid 1, the identification 1, the slope1 and the mid1 is a parameter in the first cumulative distribution model, and after 20 ten thousand sample playing time lengths are input for fitting, each of the raid 1, the identification 1, the slope1 and the mid1 can be determined.

Step 203: and fitting to obtain a second cumulative distribution model based on the long click rate by taking the sample long click rate as an input parameter.

The data processing device can obtain a preset number of sample long click rates, then draw cumulative distribution discrete data of the sample long click rates by taking the sample long click rates as input parameters, then fit the cumulative distribution discrete data to obtain a cumulative distribution curve, and determine a cumulative distribution function corresponding to the cumulative distribution curve, wherein the cumulative distribution function is a second cumulative distribution model based on the long click rates.

For example, the data processing apparatus may fit a second cumulative distribution model sigmod2 based on the long click rate as shown in the following formula (2) with a long click rate of 20 ten thousand samples as an input parameter.

In the above formula (2), y is the long click rate of the second video to be input, and each of the run 2, the main 2, the slope2 and the mid2 is a parameter in the second cumulative distribution model, and after inputting the long click rate of 20 ten thousand samples for fitting, each of the run 2, the main 2, the slope2 and the mid2 can be determined.

Step 204: and acquiring the preset number of sample playing times and the preset number of sample long clicking times.

The data processing device may obtain a preset number of sample play times as a fitting input parameter of a third cumulative distribution model based on the play times, and may obtain a preset number of sample long click times as a fitting input parameter of a fourth cumulative distribution model based on the long click times. In practical application, the sample playing times and the sample long click times can be randomly sampled from historical data of a video website by adopting a random sampling method.

For example, the data processing device may randomly sample 20 ten thousand sample play times and 20 ten thousand sample long click times from the historical data of the video website.

Step 205: and fitting to obtain a third cumulative distribution model based on the playing times by taking the playing times of the samples as input parameters.

The data processing device may obtain a preset number of sample playing times, and then draw cumulative distribution discrete data of the sample playing times, that is, cumulative distribution real data of the sample playing times, with the sample playing times as an input parameter, and then may fit the cumulative distribution discrete data, so as to obtain a cumulative distribution curve, and may determine a cumulative distribution function corresponding to the cumulative distribution curve, where the cumulative distribution function is a third cumulative distribution model based on the playing times.

Fig. 3 shows a cumulative distribution curve based on the sample playing times, and referring to fig. 3, a dotted line is cumulative distribution real data of the sample playing times, a solid line is a cumulative distribution curve obtained by fitting the cumulative distribution real data, and a cumulative distribution function corresponding to the cumulative distribution curve is a third cumulative distribution model based on the playing times.

For example, the data processing apparatus may fit a third cumulative distribution model sigmod3 based on the number of playbacks as shown in the following formula (3) with the number of playbacks of 20 ten thousand samples as an input parameter.

In the above formula (3), m is the median of the playing time lengths of the second video to be input, and each of the rain 3, the main 3, the slope3 and the mid3 is a parameter in the third cumulative distribution model, and after 20 ten thousand sample playing times are input for fitting, all of the rain 3, the main 3, the slope3 and the mid3 can be determined.

Step 206: and fitting to obtain a fourth cumulative distribution model based on the long click times by taking the sample long click times as input parameters.

The data processing device may obtain a preset number of sample long clicks, draw cumulative distribution discrete data of the sample long clicks with the sample long clicks as an input parameter, then fit the cumulative distribution discrete data to obtain a cumulative distribution curve, and determine a cumulative distribution function corresponding to the cumulative distribution curve, where the cumulative distribution function is a fourth cumulative distribution model based on the long clicks.

For example, the data processing apparatus may fit a fourth cumulative distribution model sigmod4 based on the number of long hits as shown in the following formula (4) with a long hit number of 20 ten thousand samples as an input parameter.

In the above formula (4), n is the number of long clicks of the second video to be input, and each of the raise4, the identify 4, the slope4 and the mid4 is a parameter in the fourth cumulative distribution model, and after the 20 ten thousand sample long clicks are input for fitting, each of the raise4, the identify 4, the slope4 and the mid4 can be determined.

Step 207: determining a play time median and a long click rate of a video pair in a preset time period based on a video pair consisting of a first video and a second video; the second video is a recommended video corresponding to the first video during playing and is played adjacent to the first video within a preset time period.

This step may be implemented in a manner comprising: determining a median value of each playing time length of the second video in a preset time period as a median value of the playing time lengths of the video pairs in the preset time period based on a video pair formed by the first video corresponding to the second video; determining the data number of the target playing time length which is greater than the preset playing time length or the ratio of the total time length of the second video to the preset ratio from all the playing time lengths; and determining the ratio of the data number of the target playing time length to the display number of the second video as the long click rate of the video pairs in the preset time period.

The playing time length is greater than the preset playing time length, or the ratio of the playing time length to the total time length of the second video is greater than the preset ratio, that is, one long click is realized, and correspondingly, the data number of the target playing time length, that is, the playing frequency of the second video being long clicked, is realized.

For example, the preset time period may be 7 days before the current time, and based on a video pair composed of a first video and a second video, the data processing device may determine that the median value of the playing time lengths of the video pair in 7 days is x, and the long click rate y of the video pair in 7 days.

Step 208: and determining the playing times and the long click times of the video pair in a preset time period.

The video processing device may determine the playing times of the second video within the preset time period as the playing times of the video pair within the preset time period, and may determine the long click times of the second video within the preset time period as the long click times of the video pair within the preset time period. The playing times can be used as the confidence of the median of the playing time, that is, the more the playing times are, the higher the confidence of the median of the playing time is, and thus the more accurate the result of the ranking calibration is. Similarly, the long click times can be used as the confidence of the long click rate, that is, the greater the long click times, the higher the confidence of the long click rate, and thus the more accurate the result of the ranking calibration.

For example, the data processing device may determine that the number of plays of the video pair in 7 days is m, and the number of long clicks of the video pair in 7 days is n.

Step 209: and determining a first calibration parameter of the video pair based on the playing time length through the first cumulative distribution model and the third cumulative distribution model according to the playing time length median and the playing times, and determining a second calibration parameter of the video pair based on the long click rate through the second cumulative distribution model and the fourth cumulative distribution model according to the long click rate and the long click times.

The data processing apparatus may input the median play-out time length of the video pair into the first cumulative distribution model, so that the first cumulative distribution model may output the first parameter. The data processing apparatus may input the number of plays of the pair of videos into the third cumulative distribution model, so that the third cumulative distribution model may output the third parameter. The data processing device can multiply the first parameter with the third parameter, so that a first calibration parameter of the video pair based on the playing time length can be obtained.

Likewise, the data processing device may input the long click rate of the video pair into the second cumulative distribution model, so that the second cumulative distribution model may output the second parameter. The data processing device may input the long clicks of the video pair into a fourth cumulative distribution model, so that the fourth cumulative distribution model may output a fourth parameter. The data processing device may multiply the second parameter by the fourth parameter, such that a second calibration parameter for the video pair based on the long click rate may be obtained.

It should be noted that the data processing device may determine only the first parameter and the second parameter, so that the first parameter may be determined as a first calibration parameter of the video pair based on the playing time length, and the second parameter may be determined as a second calibration parameter of the video pair based on the long click rate. Of course, in practical applications, the data processing device may also determine the first parameter, the second parameter, and the third parameter, so that the product of the first parameter and the third parameter may be determined as the first calibration parameter of the video pair based on the playing time length, and the second parameter may be determined as the second calibration parameter of the video pair based on the long click rate. In addition, the data processing device can also determine a first parameter, a second parameter and a fourth parameter, so that the first parameter can be determined as a first calibration parameter of the video pair based on the playing time length, and the product of the second parameter and the fourth parameter can be determined as a second calibration parameter of the video pair based on the long click rate. Furthermore, the data processing device may further determine the first parameter, the second parameter, the third parameter, and the fourth parameter as described in this step, so that a product of the first parameter and the third parameter may be determined as a first calibration parameter of the video pair based on the playing time length, and a product of the second parameter and the fourth parameter may be determined as a second calibration parameter of the video pair based on the long click rate, which is not specifically limited in this embodiment of the present invention.

For example, the data processing apparatus may input the median value x of the play time length of the pair of videos into the first cumulative distribution model sigmod1, so that the first cumulative distribution model sigmod1 may output the first parameter sigmod1 (x). The data processing apparatus may input the number of playcount m of the pair of videos into the third cumulative distribution model sigmod3, so that the third cumulative distribution model sigmod3 may output the third parameter sigmod3 (m). The data processing device can multiply the first parameter sigmod1(x) with the third parameter sigmod3(m), so that the first calibration parameter f of the video pair based on the playing time length can be obtained₁。

Likewise, the data processing apparatus may input the long click rate y of the pair of videos into the second cumulative distribution model sigmod2, so that the second cumulative distribution model sigmod2 may output the second parameter sigmod2 (y). The data processing apparatus may input the long number of clicks n of the video pair into the fourth cumulative distribution model sigmod4, so that the fourth cumulative distribution model sigmod4 may output a fourth parameter sigmod4 (n). The data processing device may multiply the second parameter sigmod2(y) with the fourth parameter sigmod4(n), so that the second calibration parameter f of the video pair based on the long click rate may be obtained₂。

Step 210: when the second video is included in the initial recommendation list established for the first video, the initial ranking parameter of the second video is determined.

It should be noted that the step of establishing the initial recommendation list for the first video may be performed before step 201, that is, before the whole process of the ranking calibration, and of course, in practical applications, the step of establishing the initial recommendation list for the first video may also be performed before step 210, that is, may be established before data of the initial recommendation list is needed, which is not specifically limited in this embodiment of the present invention.

In addition, since the latest user viewing data is considered in the process of establishing the initial recommendation list for the first video, the initial recommendation list obtained at this moment is different from the historical recommendation list before the first video, and therefore, the second video belongs to any one of the historical recommendation lists, but does not necessarily belong to the current initial recommendation list. When the initial recommendation list established for the first video includes the second video, the subsequent steps may be performed to calibrate the current sorting parameter of the second video using the historical play data of the second video, and when the initial recommendation list established for the first video does not include the second video, the operation may be stopped because the current sorting parameter of the other videos cannot be calibrated using the historical play data of the second video.

For example, when a second video is included in the initial recommendation list established for the first video, the data processing apparatus may determine that the initial ranking parameter of the second video is F₀。

Step 211: and calibrating the initial sequencing 5 parameters of the second video according to the first calibration parameters and the second calibration parameters to obtain the calibration sequencing parameters of the second video.

The data processing device may calibrate the initial sorting parameter of the second video by using the first calibration parameter based on the playing time length and the second calibration parameter based on the long click rate through the following formula (5), so as to obtain the calibrated sorting parameter of the second video.

In the above equation (5), F is the calibrated ranking parameter of the second video, F₀For the initial ordering parameter of the second video, f₁Is a first calibration parameter, alpha is a weight coefficient of the first calibration parameter, f₂β is a weight coefficient of the second calibration parameter.

It should be noted that, in practical applications, the weight coefficient of the first calibration parameter and the weight coefficient of the second calibration parameter may be adjusted according to practical situations, for example, when the playing duration is increased and the long click rate is increased

Also important, α may be set to 0.5 and β may be set to 0.5, the value of α may be adjusted up and the value of β may be adjusted down when increasing the playing time is more important than increasing the long click rate, and the value of α may be adjusted down and the value of β may be adjusted up when increasing the long click rate is more important than increasing the playing time.

For example, the data processing apparatus may utilize a first calibration parameter f based on the play-out duration₁And a second calibration parameter f based on the long click rate₂The initial ranking parameter F of the second video by the following equation (6)₀And calibrating to obtain a calibration sequencing parameter F of the second video.

F＝F₀×(1+0.5×f₁+0.5×f₂) (6)

Step 212: and adjusting the position of the second video in the initial recommendation list according to the calibrated sorting parameter of the second video.

For example, the data processing device may adjust the position of the second video in the initial recommendation list according to the calibrated ranking parameter F of the second video, so that a recommendation list calibrated for the first video may be obtained.

The embodiment of the invention has the following advantages:

based on a video pair composed of a first video and a second video, determining a play time length median and a long click rate of the video pair in a preset time period, wherein the second video is a recommended video corresponding to the first video when playing and is played adjacent to the first video in the preset time period, then determining a first calibration parameter of the video pair based on the play time length according to the play time length median, and determining a second calibration parameter of the video pair based on the long click rate according to the long click rate, when an initial recommendation list established for the first video includes the second video, determining an initial sorting parameter of the second video, calibrating the initial sorting parameter of the second video according to the first calibration parameter and the second calibration parameter to obtain a calibrated sorting parameter of the second video, and then calibrating the sorting parameter according to the second video, the position of the second video in the initial recommendation list is adjusted. In the embodiment of the invention, the sorting parameters of the second video in the current recommendation list of the first video can be calibrated according to the playing data of the second video in the historical recommendation list of the first video, so that the position of the second video in the current recommendation list can be adjusted, the historical playing data of the playing time and the long click rate are utilized for calibration, and the click rate and the playing time of the video can be improved during video recommendation. In addition, the accuracy of sequencing calibration can be further improved by using the playing times as the confidence coefficient of the playing time and the long click times as the confidence coefficient of the long click rate, so that the click rate and the playing time of the video are further improved when the video is recommended.

It should be noted that, for simplicity of description, the method embodiments are described as a series of acts or combination of acts, but those skilled in the art will recognize that the present invention is not limited by the illustrated order of acts, as some steps may occur in other orders or concurrently in accordance with the embodiments of the present invention. Further, those skilled in the art will appreciate that the embodiments described in the specification are presently preferred and that no particular act is required to implement the invention.

EXAMPLE III

Referring to fig. 4, a block diagram of a video sequencing calibration apparatus 400 is shown, which may specifically include:

the first determining module 401 is configured to determine, based on a video pair formed by a first video and a second video, a median value of playing time lengths and a long click rate of the video pair within a preset time period; the second video is a recommended video corresponding to the first video during playing and is played adjacent to the first video within the preset time period;

a second determining module 402, configured to determine, according to the median of the playing durations, a first calibration parameter of the video pair based on the playing durations, and determine, according to the long click rate, a second calibration parameter of the video pair based on the long click rate;

a third determining module 403, configured to determine an initial ranking parameter of the second video when the second video is included in the initial recommendation list established for the first video;

a calibration module 404, configured to calibrate an initial sorting parameter of the second video according to the first calibration parameter and the second calibration parameter, so as to obtain a calibrated sorting parameter of the second video;

an adjusting module 405, configured to adjust a position of the second video in the initial recommendation list according to the calibrated ranking parameter of the second video.

The embodiment of the invention has the following advantages:

based on a video pair composed of a first video and a second video, a play time length median and a long click rate of the video pair in a preset time period can be determined through a first determination module, wherein the second video is a recommended video corresponding to the first video during playing and is played adjacent to the first video in the preset time period, then a first calibration parameter of the video pair based on the play time length can be determined through a second determination module according to the play time length median, a second calibration parameter of the video pair based on the long click rate can be determined according to the long click rate, when an initial recommendation list established for the first video comprises the second video, an initial sorting parameter of the second video can be determined through a third determination module, and the initial sorting parameter of the second video can be calibrated through a calibration module according to the first calibration parameter and the second calibration parameter to obtain a calibrated sorting parameter of the second video, the position of the second video in the initial recommendation list may then be adjusted by the adjustment module according to the calibrated ranking parameter of the second video. In the embodiment of the invention, the sorting parameters of the second video in the current recommendation list of the first video can be calibrated according to the playing data of the second video in the historical recommendation list of the first video, so that the position of the second video in the current recommendation list can be adjusted, the historical playing data of the playing time and the long click rate are utilized for calibration, and the click rate and the playing time of the video can be improved during video recommendation.

Example four

Referring to fig. 5, there is shown a block diagram of another video sequencing calibration apparatus 500, which may specifically include;

a first determining module 501, configured to determine, based on a video pair composed of a first video and a second video, a median play time length and a long click rate of the video pair within a preset time period; the second video is a recommended video corresponding to the first video during playing and is played adjacent to the first video within the preset time period;

a second determining module 502, configured to determine, according to the median of the playing durations, a first calibration parameter of the video pair based on the playing durations, and determine, according to the long click rate, a second calibration parameter of the video pair based on the long click rate;

a third determining module 503, configured to determine an initial ranking parameter of the second video when the second video is included in the initial recommendation list established for the first video;

a calibration module 504, configured to calibrate an initial sorting parameter of the second video according to the first calibration parameter and the second calibration parameter, so as to obtain a calibrated sorting parameter of the second video;

an adjusting module 505, configured to adjust a position of the second video in the initial recommendation list according to the calibrated ranking parameter of the second video.

Optionally, referring to fig. 5, the apparatus 500 further includes:

a fourth determining module 506, configured to determine the playing times and the long click times of the video pair in the preset time period;

accordingly, the second determining module 502 comprises:

the first determining submodule 5021 is configured to determine a first calibration parameter of the video pair based on the playing time length according to the playing time length median and the playing times, and determine a second calibration parameter of the video pair based on the long click rate according to the long click rate and the long click times.

Optionally, referring to fig. 5, the apparatus 500 further includes:

a first obtaining module 507, configured to obtain a preset number of sample playing durations and sample long click rates;

a first fitting module 508, configured to fit the sample playing duration as an input parameter to obtain a first cumulative distribution model based on the playing duration;

a second fitting module 509, configured to fit the sample long click rate as an input parameter to obtain a second cumulative distribution model based on the long click rate;

accordingly, the second determining module 502 comprises:

the second determining sub-module 5022 is configured to determine, according to the median of the playing durations, a first calibration parameter of the video pair based on the playing durations through the first cumulative distribution model, and determine, according to the long click rate, a second calibration parameter of the video pair based on the long click rate through the second cumulative distribution model.

Optionally, referring to fig. 5, the apparatus 500 further includes:

a second obtaining module 510, configured to obtain the preset number of sample playing times and sample long-click times;

a third fitting module 511, configured to fit to obtain a third cumulative distribution model based on the playing times by using the sample playing times as an input parameter;

a fourth fitting module 512, configured to fit the sample long click times as an input parameter to obtain a fourth cumulative distribution model based on the long click times;

accordingly, the second determining sub-module 5022 includes:

Optionally, referring to fig. 5, the first determining module 501 includes:

the third determining submodule 5011 is configured to determine, based on a video pair formed by a first video and a second video, a median of playing time durations of the second video in a preset time period as a median of playing time durations of the video pair in the preset time period;

a fourth determining submodule 5012, configured to determine, from the respective play time durations, a data number of a target play time duration which is greater than a preset play time duration or a ratio of the total time duration of the second video to the preset ratio;

the fifth determining submodule 5013 is configured to determine a ratio of the data number of the target playing time to the display number of the second video as the long click rate of the video pair in the preset time period.

The embodiment of the invention has the following advantages:

based on a video pair composed of a first video and a second video, a play time length median and a long click rate of the video pair in a preset time period can be determined through a first determination module, wherein the second video is a recommended video corresponding to the first video during playing and is played adjacent to the first video in the preset time period, then a first calibration parameter of the video pair based on the play time length can be determined through a second determination module according to the play time length median, a second calibration parameter of the video pair based on the long click rate can be determined according to the long click rate, when an initial recommendation list established for the first video comprises the second video, an initial sorting parameter of the second video can be determined through a third determination module, and the initial sorting parameter of the second video can be calibrated through a calibration module according to the first calibration parameter and the second calibration parameter to obtain a calibrated sorting parameter of the second video, the position of the second video in the initial recommendation list may then be adjusted by the adjustment module according to the calibrated ranking parameter of the second video. In the embodiment of the invention, the sorting parameters of the second video in the current recommendation list of the first video can be calibrated according to the playing data of the second video in the historical recommendation list of the first video, so that the position of the second video in the current recommendation list can be adjusted, the historical playing data of the playing time and the long click rate are utilized for calibration, and the click rate and the playing time of the video can be improved during video recommendation. In addition, the accuracy of sequencing calibration can be further improved by using the playing times as the confidence coefficient of the playing time and the long click times as the confidence coefficient of the long click rate, so that the click rate and the playing time of the video are further improved when the video is recommended.

The embodiment of the present invention further provides a computer-readable storage medium, where a computer program is stored on the computer-readable storage medium, and when the computer program is executed by a processor, the computer program implements each process of the video sequencing calibration method embodiment, and can achieve the same technical effect, and in order to avoid repetition, details are not repeated here. The computer-readable storage medium may be a Read-only Memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk.

For the device embodiment, since it is basically similar to the method embodiment, the description is simple, and for the relevant points, refer to the partial description of the method embodiment.

The embodiments in the present specification are described in a progressive manner, each embodiment focuses on differences from other embodiments, and the same and similar parts among the embodiments are referred to each other.

As will be appreciated by one skilled in the art, embodiments of the present invention may be provided as a method, apparatus, or computer program product. Accordingly, embodiments of the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, embodiments of the present invention may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.

In a typical configuration, the computer device includes one or more processors (CPUs), input/output interfaces, network interfaces, and memory. The memory may include forms of volatile memory in a computer readable medium, Random Access Memory (RAM) and/or non-volatile memory, such as Read Only Memory (ROM) or flash memory (flash RAM). Memory is an example of a computer-readable medium. Computer-readable media, including both non-transitory and non-transitory, removable and non-removable media, may implement information storage by any method or technology. The information may be computer readable instructions, data structures, modules of a program, or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), Static Random Access Memory (SRAM), Dynamic Random Access Memory (DRAM), other types of Random Access Memory (RAM), Read Only Memory (ROM), Electrically Erasable Programmable Read Only Memory (EEPROM), flash memory or other memory technology, compact disc read only memory (CD-ROM), Digital Versatile Discs (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices, or any other non-transmission medium that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include non-transitory computer readable media (fransitory media), such as modulated data signals and carrier waves.

Embodiments of the present invention are described with reference to flowchart illustrations and/or block diagrams of methods, terminal devices (systems), and computer program products according to embodiments of the invention. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing terminal to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing terminal, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.

These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing terminal to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.

These computer program instructions may also be loaded onto a computer or other programmable data processing terminal to cause a series of operational steps to be performed on the computer or other programmable terminal to produce a computer implemented process such that the instructions which execute on the computer or other programmable terminal provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.

While preferred embodiments of the present invention have been described, additional variations and modifications of these embodiments may occur to those skilled in the art once they learn of the basic inventive concepts. Therefore, it is intended that the appended claims be interpreted as including preferred embodiments and all such alterations and modifications as fall within the scope of the embodiments of the invention.

Finally, it should also be noted that, herein, relational terms such as first and second, and the like may be used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Also, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or terminal that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or terminal. Without further limitation, an element defined by the phrase "comprising an … …" does not exclude the presence of other like elements in a process, method, article, or terminal that comprises the element.

The video sequencing calibration method and the video sequencing calibration device provided by the invention are described in detail, and the principle and the implementation mode of the invention are explained by applying specific examples, and the description of the embodiments is only used for helping to understand the method and the core idea of the invention; meanwhile, for a person skilled in the art, according to the idea of the present invention, there may be variations in the specific embodiments and the application scope, and in summary, the content of the present specification should not be construed as a limitation to the present invention.

Claims

1. A method for video sequencing calibration, the method comprising:

determining a play time median and a long click rate of a video pair in a preset time period based on the video pair formed by a first video corresponding to a second video; the second video is a recommended video corresponding to the first video during playing and is played adjacent to the first video within the preset time period; the long click rate is the ratio of the number of long clicks to the number of displays; wherein, one click when the video playing time length reaches the set proportion of the total video time length is a long click;

when the second video is included in the initial recommendation list established for the first video, determining an initial ranking parameter of the second video; the initial recommendation list is obtained by scoring a plurality of recalled videos recalled by a first video to obtain an initial sequencing parameter of each recalled video and sequencing the recalled videos according to the sequence of the initial sequencing parameters from large to small;

2. The method of claim 1, wherein before determining the first calibration parameter based on the playing time length for the video pair according to the median value of the playing time lengths and determining the second calibration parameter based on the long click rate for the video pair according to the long click rate, the method further comprises:

3. The method according to claim 2, wherein before determining the median play time length and the long click rate of the video pair in the preset time period based on the video pair composed of the first video and the second video, the method further comprises:

4. The method of claim 3, wherein before determining the median play time length and the long click rate of the video pair in the preset time period based on a video pair composed of a first video and a second video, the method further comprises:

5. The method of claim 1, wherein determining a median play time length and a long click rate of the video pair in a preset time period based on a video pair composed of a first video and a second video comprises:

6. A video sequencing calibration apparatus, the apparatus comprising:

the first determining module is used for determining a play time median and a long click rate of a video pair in a preset time period based on the video pair formed by a first video corresponding to a second video; the second video is a recommended video corresponding to the first video during playing and is played adjacent to the first video within the preset time period; the long click rate is the ratio of the number of long clicks to the number of displays; wherein, one click when the video playing time length reaches the set proportion of the total video time length is a long click;

a third determining module, configured to determine an initial ranking parameter of the second video when the second video is included in an initial recommendation list established for the first video; the initial recommendation list is obtained by scoring a plurality of recalled videos recalled by a first video to obtain an initial sequencing parameter of each recalled video and sequencing the recalled videos according to the sequence of the initial sequencing parameters from large to small;

7. The apparatus of claim 6, further comprising:

accordingly, the second determining module comprises:

8. The apparatus of claim 7, further comprising:

accordingly, the second determining module comprises:

9. The apparatus of claim 8, further comprising:

correspondingly, the second determining submodule includes:

10. The apparatus of claim 6, wherein the first determining module comprises: