Video content recommendation method and apparatus, and computer device
This application provides techniques of improving video recommendation. The techniques comprise temporarily pre-adding a first video to a recommendation video sequence; capturing a sub-sequence including the first video; modifying an initial recommendation score of the first video based on a position sequence number of the first video in the sub-sequence and a position sequence number of another video in the sub-sequence to obtain a modified recommendation score of the first video, the another video sharing a target attribute with the first video; and adding a video with a highest modified recommendation score to the recommendation video sequence.
1. A computer-implemented method of performing online video recommendation, comprising:
retrieving, by a computing device, a set of to-be-selected videos from a video library stored on a video server, the set of to-be-selected videos comprising videos based on video playback interest information determined for a first user and videos predetermined as popular based on viewing activities of a plurality of users;
consecutively selecting by the computing device a first video from the set of to-be-selected videos and temporarily pre-adding the first video to a last position of a sequence of recommendation videos;
capturing, by the computing device, from the sequence of recommendation videos, a sub-sequence comprising a preset quantity of videos, wherein the sub-sequence comprises the first video;
obtaining, by the computing device, a target attribute and an initial recommendation score of the first video;
identifying, by the computing device, at least one target video in the sub-sequence, wherein the at least one target video has the target attribute and is different from the first video, and determining a quantity n of the at least one target video;
modifying, by the computing device, the initial recommendation score of the first video based on the quantity n and a position distance between each of the at least one target video and the first video to obtain a modified recommendation score, wherein the modifying the initial recommendation score of the first video based on the quantity n and a position distance between each of the at least one target video and the first video to obtain a modified recommendation score further comprises:
obtaining a position sequence number i of the at least one target video in the sub-sequence and a position sequence number k of the first video in the sub-sequence in response to determining that the quantity n is less than a preset threshold N, and
modifying the initial recommendation score based on the position sequence number i of the at least one target video, the position sequence number k of the first video, and a preset modification formula, wherein the preset modification formula comprises:
score
k
′
=
score
k
count
(
tags
k
)
*
(
∑
tag
=
1
count
(
tags
k
)
∏
i
=
0
k
-
1
(
1
-
1
count
(
tags
i
)
*
demote
(
distance
(
i
,
k
)
,
tag
)
)
*
∏
i
=
0
k
-
1
(
1
-
demote
(
distance
(
i
,
k
)
,
up
)
)
wherein score k represents the initial recommendation score of the first video, score′ k represents the modified recommendation score of the first video, count(tags k ) represents a quantity of tag attributes associated with the first video, count(tags i ) represents a quantity of tag attributes associated with the target video, demote(distance(i, k), tag) represents a first modification function associated with tag attributes, demote(distance(i, k), up) represents a second modification function associated with uploader attributes indicating information about users who uploads videos:
obtaining, by the computing device, modified recommendation scores of all videos in the set of to-be-selected videos; and
selecting, by the computing device, a video with a highest modified recommendation score from the set of to-be-selected videos and adding the video with the highest modified recommendation score to the last position of the sequence of recommendation videos.
2. The method according to claim 1 , wherein the obtaining a target attribute of the first video comprises:
in response to determining that the target attribute of the first video is empty, marking a preset video attribute on the first video and identifying the preset video attribute as the target attribute of the first video.
3. The method according to claim 1 , wherein the modifying the initial recommendation score of the first video based on the quantity n and a position distance between each of the at least one target video and the first video to obtain a modified recommendation score further comprises:
in response to determining that the quantity n is greater than or equal to a preset threshold N, setting the modified recommendation score of the first video to 0.
4. The method according to claim 1 , wherein the first modification function and the second modification function are attenuation functions whose value is greater than or equal to 0 and less than or equal to 1, and wherein the attenuation functions comprise a linear function or a quadratic function.
5. The method according to claim 1 , wherein when both the first modification function and the second modification function are half-life functions, the modification formula comprises:
score
k
′
=
score
k
c
o
u
n
t
(
t
a
g
s
k
)
*
(
∑
tag
=
1
count
(
tags
k
)
∏
i
=
0
k
-
1
(
1
-
1
count
(
tags
i
)
(
1
2
)
distance
(
i
,
k
)
T
tag
)
)
*
∏
i
=
0
k
-
1
(
1
-
(
1
2
)
distance
(
i
,
k
)
T
up
)
wherein Ttag represents a preset attenuation constant corresponding to a tag attribute, and Tup represents a preset attenuation constant corresponding to an uploader attribute.
6. The method according to claim 1 , wherein the modifying the initial recommendation score of the first video based on the quantity n and a position distance between each of the at least one target video and the first video and the first video to obtain a modified recommendation score further comprises:
obtaining a target user attribute associated with a user terminal device to which the video content is recommended; and
modifying the initial recommendation score of the first video based on the quantity n and the position distance between each of the at least one target video and the first video with reference to the target user attribute to obtain a modified recommendation score corresponding to a target user.
7. A computer device of implementing online video recommendation, wherein the computer device comprises a memory and a processor, computer-readable instructions capable of running on the processor are stored in the memory, and the computer-readable instructions are executable by the processor to implement operations comprising:
accessing a set of to-be-selected videos from a video library stored on a video server, the set of to-be-selected videos comprising videos based on video playback interest information determined for a first user and videos predetermined as popular based on the viewing activities of a plurality of users;
consecutively selecting a first video from the set of to-be-selected videos and temporarily pre-adding the first video to a last position of a sequence of recommendation videos;
capturing, from the sequence of recommendation videos, a sub-sequence comprising a preset quantity of videos, wherein the sub-sequence comprises the first video;
obtaining a target attribute and an initial recommendation score of the first video;
identifying at least one target video in the sub-sequence, wherein the at least one target video has the target attribute and is different from the first video, and determining a quantity n of the at least one target video;
modifying the initial recommendation score of the first video based on the quantity n and a position distance between each of the at least one target video and the first video to obtain a modified recommendation score, wherein the modifying the initial recommendation score of the first video based on the quantity n and a position distance between each of the at least one target video and the first video to obtain a modified recommendation score further comprises:
obtaining a position sequence number i of the at least one target video in the sub-sequence and a position sequence number k of the first video in the sub-sequence in response to determining that the quantity n is less than a preset threshold N, and
modifying the initial recommendation score based on the position sequence number i of the at least one target video, the position sequence number k of the first video, and a preset modification formula, wherein the preset modification formula comprises:
score
k
′
=
score
k
count
(
tags
k
)
*
(
∑
tag
=
1
count
(
tags
k
)
∏
i
=
0
k
-
1
(
1
-
1
count
(
tags
i
)
*
demote
(
distance
(
i
,
k
)
,
tag
)
)
*
∏
i
=
0
k
-
1
(
1
-
demote
(
distance
(
i
,
k
)
,
up
)
)
wherein score k represents the initial recommendation score of the first video, score′ k represents the modified recommendation score of the first video, count(tags k ) represents a quantity of tag attributes associated with the first video, count(tags i ) represents a quantity of tag attributes associated with the target video, demote(distance(i, k), tag) represents a first modification function associated with tag attributes, demote(distance(i, k), up) represents a second modification function associated with uploader attributes indicating information about users who uploads videos;
obtaining modified recommendation scores of all videos in the set of to-be-selected videos; and
selecting a video with a highest modified recommendation score from the set of to-be-selected videos and adding the video with the highest modified recommendation score to the last position of the sequence of recommendation videos.
8. The computer device according to claim 7 , wherein the obtaining a target attribute of the first video comprises:
in response to determining that the target attribute of the first video is empty, marking a preset video attribute on the first video and identifying the preset video attribute as the target attribute of the first video.
9. The computer device according to claim 7 , wherein the first modification function and the second modification function are attenuation functions whose value is greater than or equal to 0 and less than or equal to 1, and wherein the attenuation functions comprise a linear function or a quadratic function; and wherein when both the first modification function and the second modification function are half-life functions, the modification function comprises:
score
k
′
=
score
k
c
o
u
n
t
(
t
a
g
s
k
)
*
(
∑
tag
=
1
count
(
tags
k
)
∏
i
=
0
k
-
1
(
1
-
1
count
(
tags
i
)
(
1
2
)
distance
(
i
,
k
)
T
tag
)
)
*
∏
i
=
0
k
-
1
(
1
-
(
1
2
)
distance
(
i
,
k
)
T
up
)
wherein Ttag represents a preset attenuation constant corresponding to the tag attribute, and Tup represents a preset attenuation constant corresponding to the uploader attribute.
10. The computer device according to claim 7 , wherein the modifying the initial recommendation score of the first video based on the quantity n and a position distance between each of the at least one target video and the first video to obtain a modified recommendation score further comprises:
obtaining a target user attribute associated with a user terminal device to which the video content is recommended; and
modifying the initial recommendation score of the first video based on the quantity n and the position distance between each of the at least one target video and the first video with reference to the target user attribute to obtain a modified recommendation score corresponding to a target user.
11. A non-transitory computer-readable storage medium, wherein the computer-readable storage medium stores computer-readable instructions, and the computer-readable instructions are capable of being executed by at least one processor to enable the at least one processor to perform operations comprising:
retrieving a set of to-be-selected videos from a video library stored on a video server, the set of to-be-selected videos comprising videos based on video playback interest information determined for a first user and videos predetermined as popular based on the viewing activities of a plurality of users;
consecutively selecting a first video from the set of to-be-selected videos and temporarily pre-adding the first video to a last position of a sequence of recommendation videos;
capturing, from the sequence of recommendation videos, a sub-sequence comprising a preset quantity of videos, wherein the sub-sequence comprises the first video;
obtaining a target attribute and an initial recommendation score of the first video;
identifying at least one target video in the sub-sequence, wherein the at least one target video has the target attribute and is different from the first video, and determining a quantity n of the at least one target video;
modifying the initial recommendation score of the first video based on the quantity n and a position distance between each of the at least one target video and the first video to obtain a modified recommendation score, wherein the modifying the initial recommendation score of the first video based on the quantity n and a position distance between each of the at least one target video and the first video to obtain a modified recommendation score further comprises:
obtaining a position sequence number i of the at least one target video in the sub-sequence and a position sequence number k of the first video in the sub-sequence in response to determining that the quantity n is less than a preset threshold N, and
modifying the initial recommendation score based on the position sequence number i of the at least one target video, the position sequence number k of the first video, and a preset modification formula, wherein the preset modification formula comprises:
score
k
′
=
score
k
count
(
tags
k
)
*
(
∑
tag
=
1
count
(
tags
k
)
∏
i
=
0
k
-
1
(
1
-
1
count
(
tags
i
)
*
demote
(
distance
(
i
,
k
)
,
tag
)
)
*
∏
i
=
0
k
-
1
(
1
-
demote
(
distance
(
i
,
k
)
,
up
)
)
wherein score k represents the initial recommendation score of the first video, score′ k represents the modified recommendation score of the first video, count(tags k ) represents a quantity of tag attributes associated with the first video, count(tags i ) represents a quantity of tag attributes associated with the target video, demote(distance(i, k), tag) represents a first modification function associated with tag attributes, demote(distance(i, k), up) represents a second modification function associated with uploader attributes indicating information about users who uploads videos;
obtaining modified recommendation scores of all videos in the set of to-be-selected videos;
selecting a video with a highest modified recommendation score from the set of to-be-selected videos and adding the video with the highest modified recommendation score to the last position of the sequence of recommendation videos.
12. The non-transitory computer-readable storage medium according to claim 11 , wherein the obtaining a target attribute of the first video comprises:
in response to determining that the target attribute of the first video is empty, marking a preset video attribute on the first video and identifying the preset video attribute as the target attribute of the first video.
13. The non-transitory computer-readable storage medium according to claim 11 , wherein the first modification function and the second modification function are attenuation functions whose value is greater than or equal to 0 and less than or equal to 1, and wherein the attenuation functions comprise a linear function or a quadratic function.
14. The non-transitory computer-readable storage medium according to claim 11 , wherein when both the first modification function and the second modification function are half-life functions, the modification formula comprises:
score
k
′
=
score
k
c
o
u
n
t
(
t
a
g
s
k
)
*
(
∑
tag
=
1
count
(
tags
k
)
∏
i
=
0
k
-
1
(
1
-
1
count
(
tags
i
)
(
1
2
)
distance
(
i
,
k
)
T
tag
)
)
*
∏
i
=
0
k
-
1
(
1
-
(
1
2
)
distance
(
i
,
k
)
T
up
)
wherein Ttag represents a preset attenuation constant corresponding to the tag attribute, and Tup represents a preset attenuation constant corresponding to the uploader attribute.