CN111694971A - Content recommendation method and system - Google Patents
Content recommendation method and system Download PDFInfo
- Publication number
- CN111694971A CN111694971A CN202010524482.6A CN202010524482A CN111694971A CN 111694971 A CN111694971 A CN 111694971A CN 202010524482 A CN202010524482 A CN 202010524482A CN 111694971 A CN111694971 A CN 111694971A
- Authority
- CN
- China
- Prior art keywords
- content
- user
- cluster
- cloud server
- feature vector
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 35
- 239000013598 vector Substances 0.000 claims abstract description 169
- 238000012163 sequencing technique Methods 0.000 claims abstract description 10
- 230000006399 behavior Effects 0.000 claims description 22
- 238000005192 partition Methods 0.000 claims description 2
- 238000010586 diagram Methods 0.000 description 7
- 238000001914 filtration Methods 0.000 description 6
- 230000008859 change Effects 0.000 description 2
- 230000002776 aggregation Effects 0.000 description 1
- 238000004220 aggregation Methods 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 230000004044 response Effects 0.000 description 1
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
- G06F16/40—Information retrieval; Database structures therefor; File system structures therefor of multimedia data, e.g. slideshows comprising image and additional audio data
- G06F16/43—Querying
- G06F16/432—Query formulation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/907—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/40—Information retrieval; Database structures therefor; File system structures therefor of multimedia data, e.g. slideshows comprising image and additional audio data
- G06F16/43—Querying
- G06F16/435—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
- G06F16/40—Information retrieval; Database structures therefor; File system structures therefor of multimedia data, e.g. slideshows comprising image and additional audio data
- G06F16/44—Browsing; Visualisation therefor
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/40—Information retrieval; Database structures therefor; File system structures therefor of multimedia data, e.g. slideshows comprising image and additional audio data
- G06F16/45—Clustering; Classification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9532—Query formulation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/958—Organisation or management of web site content, e.g. publishing, maintaining pages or automatic linking
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Databases & Information Systems (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Multimedia (AREA)
- Mathematical Physics (AREA)
- Library & Information Science (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
- Information Transfer Between Computers (AREA)
- Two-Way Televisions, Distribution Of Moving Picture Or The Like (AREA)
Abstract
The application discloses a content recommendation method, which comprises the following steps: the intelligent terminal generates a score of the watching content according to the behavior of the user on the watching content and sends the score to the edge server; the edge server updates the interest characteristic vector of the user and sends the score and the updated interest characteristic vector of the user to the cloud server; when the cloud server needs to recommend network content for the user, determining a user cluster which is most matched with the user in all user clusters of the cloud server by using the received updated interest characteristic vector of the user, and adding the content corresponding to the user cluster into an initial recommendation queue; and in the initial recommendation queue, deleting the content of the behavior generated by the user, sequencing the updated initial recommendation queue, and recommending the first N contents to the user. By the application, the computing pressure of the cloud server can be greatly reduced.
Description
Technical Field
The present application relates to computer technologies, and in particular, to a method and a system for recommending content.
Background
With the rise of content platforms such as short videos and news information, users like to browse various short videos and news apps, such as judder, micro-video, today's headline, etc., through clients at leisure time. When different users browse the same content platform, the platform needs to push different contents due to different points of interest. In addition, for the user, his interest is often constantly changing, and due to the short video and news viewing time, if the server cannot quickly respond to the change of the user's interest in a short time, the user's experience is poor.
Recommendation systems in different industries are similar in nature although the application fields and scenes are different. All recommendation systems need to rely on characteristics of both content and user interest to decide on specific recommended content. At present, all features (content features and user interest features) are calculated at a cloud server side, and with the increase of the number of users, the computing pressure of the cloud server is increased. With the annual increase of the network speed and the continuous increase of the user demand, the adoption of the existing content recommendation method can bring greater and greater computing pressure to the cloud server, and how to reduce the computing pressure of the cloud server becomes a problem which needs to be solved urgently in a new era.
Disclosure of Invention
The application provides a content recommendation method and system, which can greatly reduce the computing pressure of a cloud server.
In order to achieve the purpose, the following technical scheme is adopted in the application:
a method of recommending content, comprising:
the intelligent terminal generates a score of the watching content according to the behavior of the user on the watching content and sends the score to the edge server; wherein the score is used for representing the interest degree of the user in the viewing content;
the edge server receives the score, determines an interest characteristic vector of the user, and queries a content characteristic vector of the watching content from a cloud server; updating the interest feature vector of the user according to the interest feature vector of the user, the content feature vector of the watching content and the score, and sending the score and the updated interest feature vector of the user to the cloud server;
when the cloud server needs to recommend network content for the user, determining a user cluster which is most matched with the user in all user clusters of the cloud server by using the received updated interest characteristic vector of the user, and adding the content corresponding to the user cluster into an initial recommendation queue;
deleting the content of the behavior generated by the user in an initial recommendation queue, sequencing the updated initial recommendation queue, and recommending the first N contents to the user; wherein, N is a preset positive integer.
Preferably, after the cloud server receives the score and the updated interest feature vector of the user, the method further includes:
the cloud server updates the content feature vector of the watching content according to the received score, the content feature vector of the watching content and the interest feature vector of the user before/after updating;
and when the cloud server determines that the user is interested in the watching content according to the score, determining a content cluster which is most matched with the watching content in all content clusters of the cloud server by using the updated content feature vector of the watching content, and adding the content corresponding to the content cluster into an initial recommendation queue.
Preferably, the determining a user cluster that is most matched with the user in all the user clusters of the cloud server includes: and calculating the distance between the user and the center of each user cluster by using the updated interest characteristic vector of the user, and finding out the user cluster with the closest distance as the user cluster which is most matched with the user.
Preferably, the determining a content cluster that best matches the viewing content from among all content clusters of the cloud server includes: and calculating the distance between the watching content and the center of each content cluster by using the updated content feature vector of the watching content, and finding out the content cluster with the closest distance as the content cluster which is most matched with the watching content.
Preferably, when the user is interested in the viewing content, the method further comprises: and adding the watching content into a content queue corresponding to the user cluster which is most matched with the user.
Preferably, after updating the content feature vector of the viewing content, the method further comprises: and removing the watching content from the current content cluster, and adding the content cluster which is most matched with the watching content.
Preferably, the manner of dividing the user cluster by the cloud server includes:
the cloud server divides all users into different clusters by using a k-means algorithm according to the interest feature vectors of all users, and the difference between the interest feature vectors of different users in each cluster is within a first set range.
Preferably, the manner of dividing the content cluster by the cloud server includes:
and the cloud server divides all the contents into different clusters by using a k-means algorithm according to the content feature vectors of all the contents, and the difference between the content feature vectors of different contents in each cluster is within a second set range.
Preferably, the determining, by the edge server, the interest feature vector of the user after receiving the score includes: the edge server inquires the interest feature vector of the user locally, and if the interest feature vector of the user is not inquired, the edge server acquires the interest feature vector of the user from the cloud server.
A system for recommending contents, comprising: the system comprises an intelligent terminal, an edge server and a cloud server;
the intelligent terminal is used for generating a score of the watching content according to the behavior of the user on the watching content and sending the score to the edge server; wherein the score is used for representing the interest degree of the user in the viewing content;
the edge server is used for determining the interest characteristic vector of the user after receiving the score, and inquiring the content characteristic vector of the watching content from the cloud server; updating the interest feature vector of the user according to the interest feature vector of the user, the content feature vector of the watching content and the score, and sending the score and the updated interest feature vector of the user to the cloud server;
the cloud server is used for determining a user cluster which is closest to the user in all user clusters of the cloud server by using the received updated interest characteristic vector of the user when the network content needs to be recommended to the user, and adding the content corresponding to the user cluster into an initial recommendation queue; the intelligent terminal is also used for deleting the content of the behavior generated by the user in the initial recommendation queue, sequencing the updated initial recommendation queue and recommending the first N contents to the user through the intelligent terminal; wherein, N is a preset positive integer.
Preferably, the cloud server includes: the device comprises a receiving unit, a learning unit, a storage unit and a recommendation engine unit;
the receiving unit is used for receiving the score and the updated interest feature vector of the user;
the learning unit is used for updating the content feature vector of the watching content according to the score, the content feature vector of the watching content stored in the storage unit and the interest feature vector of the user before/after updating, and storing the updated content feature vector of the watching content in the storage unit;
the storage unit is used for storing interest feature vectors of all users and content feature vectors of all contents;
the recommendation engine unit is used for determining a user cluster which is closest to the user in all user clusters of the cloud server by using the received updated interest feature vector of the user when the network content needs to be recommended for the user, and adding the content corresponding to the user cluster into an initial recommendation queue; the cloud server is further used for determining a content cluster which is closest to the watching content in all content clusters of the cloud server by using the updated content feature vector of the watching content when the user is determined to be interested in the watching content according to the score, and adding the content corresponding to the content cluster into an initial recommendation queue; and the system is also used for deleting the content of the behavior generated by the user in the initial recommendation queue, sequencing the updated initial recommendation queue and recommending the first N contents to the user.
According to the technical scheme, the intelligent terminal generates the score of the watching content according to the behavior of the user on the watching content and sends the score to the edge server; the edge server determines the interest characteristic vector of the user after receiving the score, and inquires the content characteristic vector of the watching content from the cloud server; updating the interest characteristic vector of the user according to the interest characteristic vector of the user, the content characteristic vector of the watching content and the score, and sending the score and the updated interest characteristic vector of the user to the cloud server; when the cloud server needs to recommend network content for a user, determining a user cluster which is most matched with the user in all user clusters of the cloud server by using the received updated interest characteristic vector of the user, and adding the content corresponding to the user cluster into an initial recommendation queue; and in the initial recommendation queue, deleting the content of which the user has generated behavior, sequencing the updated initial recommendation queue, and recommending the first N contents to the user. Through the processing, the updating processing of the user interest feature vector is completed on the edge server, and the content feature vector is updated and the recommended content is matched and selected on the cloud server, so that the computing pressure of the cloud server is greatly reduced.
Drawings
Fig. 1 is a schematic diagram of a basic flow of a content recommendation method in the present application;
FIG. 2 is a schematic diagram of a basic structure of a content recommendation system according to the present application;
fig. 3 is a processing flow chart of the intelligent terminal in the content recommendation method of the present application;
FIG. 4 is a flowchart illustrating a process of an edge server in the content recommendation method of the present application;
fig. 5 is a schematic processing diagram of the content recommendation method in the present application in a cloud server;
FIG. 6 is a flowchart of a recommended content in a cloud server according to the method of the present application;
FIG. 7 is a schematic diagram of user cluster associated content in a cloud server according to the method of the present application
Fig. 8 is a schematic diagram of a content cluster in a cloud server according to the method of the present application.
Detailed Description
For the purpose of making the objects, technical means and advantages of the present application more apparent, the present application will be described in further detail with reference to the accompanying drawings.
The content recommendation method and the content recommendation system provided by the application provide a new content recommendation mode through the cooperation of the edge server and the cloud server. The method comprises the steps that the interest characteristics of a user are calculated through the edge server, the characteristics of the content are calculated through the cloud server, personalized content can be provided for the user through operations such as aggregation, filtering and sequencing, the pressure of the cloud server end is relieved through the mode that the interest characteristics of the user are calculated and transferred to the edge server, the interest change of the user is responded quickly, and the content which the user is interested in is pushed in time.
Fig. 1 is a schematic diagram of a basic flow of a content recommendation method in the present application, and fig. 2 is a schematic diagram of a basic structure of a content recommendation system in the present application. Fig. 1 may be implemented in the system shown in fig. 2. As shown in fig. 1 and 2, the method includes:
The processing of this step is completed at the intelligent terminal. When the user A watches the content B (such as a short video), the intelligent terminal collects the behavior of the user A on watching the content B and generates a score according to the behavior. Wherein, the score is used to indicate the interest degree of the user a in the viewing content B, i.e. the favorite degree of the user a in the viewing content B, and preferably, the larger the score is, the more the user a likes the viewing content B.
The processing flow of the intelligent terminal can be as shown in fig. 3. The specific intelligent terminal can comprise the following four modules:
1. receiving unit
The unit is mainly used for receiving the content pushed to the client by the cloud server from the cloud server.
2. Behavior acquisition unit
The unit is mainly used for collecting user behaviors, such as whether to watch repeatedly, whether to approve, whether to forward, whether to leave after only seeing one or two seconds, and the like.
3. Behavior conversion unit
The unit converts the user's behavior into a rating score, generates the user's preference for the content, and indicates that the user likes the more the score is larger.
4. Transmitting unit
The unit is for the client to transmit the content score to the edge server.
And step 102, after receiving the score, the edge server determines the interest characteristic vector of the user A, and inquires the content characteristic vector of the watching content B from the cloud server.
The interest feature vector of the user is used for representing the interest feature of the user, and can comprise a plurality of aspects of interests, so that the interest feature vector is represented in a vector form. The content feature vector is used to characterize a specific feature of a content, and may also include features of multiple aspects, and is thus represented in vector form. For example, when the content is a movie, the content feature vector thereof may include the movie name, the movie type, the movie lead actor, and the like.
The edge server firstly queries the interest feature vector of the user A locally, and if the interest feature vector of the user A is not queried locally, the interest feature vector of the user A is obtained from the cloud server. Further, the edge server also needs to query the content feature vector of the viewing content B from the cloud server.
And 103, the edge server updates the interest feature vector of the user A according to the inquired old interest feature vector of the user A, the content feature vector of the watching content B and the received score, and sends the score and the updated new interest feature vector of the user A to the cloud server.
In the step, the interest characteristic vector of the user A is updated at the edge server. The interest feature vector of the user A inquired by the edge server is called an old interest feature vector, and the updated interest feature vector of the user A is called a new interest feature vector. The edge server calculates the new interest feature vector of user a using the old interest feature vector of user a, the content feature vector of viewing content B and the received score. The new interest feature vector may be calculated specifically by a collaborative filtering algorithm. After the new interest feature vector is obtained through calculation, the interest feature vector and the score received by the edge server are sent to the cloud server, so that the cloud server can update and store the interest feature vector of the user.
The processing of the above steps 102 to 103 is completed in the edge server, as shown in fig. 4, a specific edge server may include the following three units:
1. receiving unit
The unit receives the score of the user A from the intelligent terminal on the watched content B, searches the old interest feature vector of the user from the local, and queries the old interest feature vector of the user from the cloud server side if the old interest feature vector of the user is not found.
The unit also queries the content feature vector of the content B seen by the user a from the cloud server.
2. Learning unit
The unit calculates a new user interest feature vector through the acquired content score, the content feature vector and the old user interest feature vector by a collaborative filtering algorithm.
3. Transmitting unit
The unit sends the calculated new interest feature vector of the user and the collected content score to the cloud server.
And 104, when the cloud server needs to recommend network content for the user A, determining a user cluster which is most matched with the user A in all the user clusters of the cloud server by using the received updated interest characteristic vector of the user A, and adding the content corresponding to the user cluster into an initial recommendation queue.
All users are divided into a plurality of user clusters in the cloud server, and the interests of different users in the same user cluster are similar. Meanwhile, all the contents are divided into a plurality of content clusters in the cloud server, and the characteristics of different contents in the same content cluster are similar.
Preferably, the manner in which the cloud server partitions the user cluster may include: the cloud server divides all users into different clusters by using a k-means algorithm according to the interest feature vectors of all users, and the difference between the interest feature vectors of different users in each user cluster is within a first set range, so that different users in each user cluster are ensured to have similar interest feature vectors. Each user cluster is correspondingly provided with a content queue for storing the content which is interested by the users in the cluster.
Preferably, the manner of dividing the content cluster by the cloud server may include: the cloud server divides all contents into different clusters by using a k-means algorithm according to the content feature vectors of all contents, and the difference between the content feature vectors of different contents in each cluster is within a second set range, so that the different contents in each content cluster are ensured to have similar content feature vectors.
When recommended content is selected for the user A, based on the user cluster, the user cluster which is most matched with the user A in all the user clusters of the cloud server can be determined by using the received updated interest feature vector of the user A, and the content corresponding to the user cluster is added into an initial recommendation queue. The determining of the user cluster most matched with the user a may be calculating a distance between the user a and a center of each user cluster by using the updated interest feature vector of the user a, and finding out the user cluster closest to the user a as the user cluster most matched with the user a. Of course, other existing matching modes can be selected, and are not described in detail here.
The initial recommendation queue, which is the simplest initial recommendation queue, can be obtained in the above manner, and then step 106 can be directly performed. Alternatively, on the basis of the above initial recommendation queue, it is preferable to further enrich the initial recommendation queue by the following step 105, and then perform step 106.
105, the cloud server updates the content feature vector of the watching content B according to the received score, the content feature vector of the watching content B and the interest feature vector of the user A before updating; when the user A is determined to be interested in the watching content B according to the score, determining a content cluster which is most matched with the watching content B in all content clusters of the cloud server by using the updated content feature vector of the watching content B, and adding the content corresponding to the content cluster into an initial recommendation queue.
Here, the cloud server first updates the content feature vector of the viewing content B after receiving the score from the edge server. Specifically, the content feature vector of the viewing content B is updated according to the received score, the content feature vector of the viewing content B, and the interest feature vector of the user a before update.
After the content feature vector is updated, if the user A is determined to be interested in the watching content B according to the score, determining a content cluster which is most matched with the watching content B in all content clusters of the cloud server by using the updated content feature vector of the watching content B, and adding the content corresponding to the content cluster into an initial recommendation queue. The determining of the content cluster that is most matched with the viewing content B may be calculating the distance between the viewing content B and the center of each content cluster by using the updated content feature vector of the viewing content B, and finding the content cluster that is closest to the viewing content B as the content cluster that is most matched with the viewing content B.
In addition, preferably, when the user a is interested in the viewing content B, the viewing content B may be further added to the content queue corresponding to the user cluster that is most matched with the user a, and the viewing content B is removed from the content cluster where the user a is currently located and added to the content cluster that is most matched with the viewing content B.
And 106, deleting the contents of the behavior generated by the user A in the initial recommendation queue, sequencing the updated initial recommendation queue, and recommending the first N contents to the user A.
The processing of this step is performed in the cloud server. Filtering the initial recommendation queue to remove the content of the behavior generated by the user A. Further, the updated initial recommendation queue may be sorted, and the specific sorting basis may be set according to needs, for example, sorting according to novelty, popularity, diversity, and the like. And finally recommending the first N contents in the initial recommendation queue to the user A, namely sending the contents to the intelligent terminal.
The processing of the steps 104-106 is completed in the cloud server, and is used for calculating an algorithm model of the content feature vector and a recommendation model, as shown in fig. 5. A particular cloud server may include the following five units:
1. receiving unit
The unit receives the user's score of the viewed content and the user's new interest feature vector from the edge server.
2. Learning unit
The unit learns new content feature vectors according to the scores of the content of the users, the user interest feature vectors and the old content feature vectors which are transmitted by the edge server and a collaborative filtering algorithm. The user interest feature vector used in updating the content feature vector can be an old user interest feature vector or a new user interest feature vector, and experiments show that the new content feature vector obtained by using the old user interest feature is more accurate.
3. Memory cell
The unit stores a user interest feature vector, and a content feature vector.
4. Recommendation engine unit
Dividing users into p clusters by using a k-means algorithm according to all user interest feature vectors, wherein each cluster has similar interest feature vectors; and dividing the content into k clusters by using a k-means algorithm according to the content feature vector, wherein the content of each cluster has similar feature vectors.
When recommending content to the user A, the unit calculates the distances between the user A and the centers of all p user clusters by using the new interest characteristic vectors of the user sent by the edge server, finds out the user cluster q closest to the user A, and puts the content corresponding to the user cluster q into an initial recommendation queue. As shown in fig. 6.
And determining whether the user A likes the content d according to the likeness degree score of the user A to the content d received by the receiving unit, if yes, calculating the distances between the content d and the centers of all k content clusters by using a new content feature vector of the content d calculated by the cloud server learning unit, finding out a content cluster e closest to the content cluster e, and adding the content corresponding to the content cluster e into an initial recommendation queue. As shown in fig. 6.
If the user A likes the content b, calculating the distance between the user A and the centers of p user clusters by using the new interest characteristic vector of the user sent by the edge server, finding out the user cluster q closest to the user A, and putting the content b into a content queue associated with the user cluster q. As shown in fig. 7.
And calculating the distance between the content d and the centers of the k content clusters by using the new characteristic vector of the content d calculated by the cloud server learning unit, finding out the content cluster e closest to the content d, removing the content d from the original content cluster, and placing the content d into a content queue associated with the content cluster e. As shown in fig. 8.
Filtering the initial recommendation queue to remove the content of the behavior generated by the user A; sorting the initial recommendation queues, and sorting the contents according to novelty, popularity and diversity; and putting the first N contents of the initial recommendation queue into a final recommendation queue.
So far, the content recommendation method flow in the present application is finished. Through the processing of the application, each intelligent terminal transmits the scores of the user on the content (such as the short videos) to the edge server, the edge server completes the learning of the new interest feature vectors of the user based on the machine learning algorithm, the cloud server does not need to calculate the interest feature vectors of the user any more, and only needs to push the new content (such as the short videos) to the user according to the recommendation algorithm and the new interest feature vectors of the user transmitted by the edge server, so that the response time is optimized.
The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents, improvements and the like made within the spirit and principle of the present invention should be included in the scope of the present invention.
Claims (11)
1. A method for recommending a content, comprising:
the intelligent terminal generates a score of the watching content according to the behavior of the user on the watching content and sends the score to the edge server; wherein the score is used for representing the interest degree of the user in the viewing content;
the edge server receives the score, determines an interest characteristic vector of the user, and queries a content characteristic vector of the watching content from a cloud server; updating the interest feature vector of the user according to the interest feature vector of the user, the content feature vector of the watching content and the score, and sending the score and the updated interest feature vector of the user to the cloud server;
when the cloud server needs to recommend network content for the user, determining a user cluster which is most matched with the user in all user clusters of the cloud server by using the received updated interest characteristic vector of the user, and adding the content corresponding to the user cluster into an initial recommendation queue;
deleting the content of the behavior generated by the user in an initial recommendation queue, sequencing the updated initial recommendation queue, and recommending the first N contents to the user; wherein, N is a preset positive integer.
2. The method of claim 1, wherein after the cloud server receives the score and the updated interest feature vector of the user, the method further comprises:
the cloud server updates the content feature vector of the watching content according to the received score, the content feature vector of the watching content and the interest feature vector of the user before/after updating;
and when the cloud server determines that the user is interested in the watching content according to the score, determining a content cluster which is most matched with the watching content in all content clusters of the cloud server by using the updated content feature vector of the watching content, and adding the content corresponding to the content cluster into an initial recommendation queue.
3. The method according to claim 1 or 2, wherein the determining a cluster of users that is the best match to the user among all clusters of users of the cloud server comprises: and calculating the distance between the user and the center of each user cluster by using the updated interest characteristic vector of the user, and finding out the user cluster with the closest distance as the user cluster which is most matched with the user.
4. The method of claim 2, wherein the determining a content cluster that best matches the viewing content from among all content clusters of the cloud server comprises: and calculating the distance between the watching content and the center of each content cluster by using the updated content feature vector of the watching content, and finding out the content cluster with the closest distance as the content cluster which is most matched with the watching content.
5. The method of claim 1 or 2, wherein when the user is interested in the viewing content, the method further comprises: and adding the watching content into a content queue corresponding to the user cluster which is most matched with the user.
6. The method of claim 2 or 4, wherein after updating the content feature vector of the viewing content, the method further comprises: and removing the watching content from the current content cluster, and adding the content cluster which is most matched with the watching content.
7. The method according to claim 1 or 2, wherein the cloud server partitions the user cluster by:
the cloud server divides all users into different clusters by using a k-means algorithm according to the interest feature vectors of all users, and the difference between the interest feature vectors of different users in each cluster is within a first set range.
8. The method according to claim 1 or 2, wherein the cloud server divides the content clusters by:
and the cloud server divides all the contents into different clusters by using a k-means algorithm according to the content feature vectors of all the contents, and the difference between the content feature vectors of different contents in each cluster is within a second set range.
9. The method of claim 1 or 2, wherein determining the interest feature vector of the user after the edge server receives the score comprises: the edge server inquires the interest feature vector of the user locally, and if the interest feature vector of the user is not inquired, the edge server acquires the interest feature vector of the user from the cloud server.
10. A system for recommending contents, comprising: the system comprises an intelligent terminal, an edge server and a cloud server;
the intelligent terminal is used for generating a score of the watching content according to the behavior of the user on the watching content and sending the score to the edge server; wherein the score is used for representing the interest degree of the user in the viewing content;
the edge server is used for determining the interest characteristic vector of the user after receiving the score, and inquiring the content characteristic vector of the watching content from the cloud server; updating the interest feature vector of the user according to the interest feature vector of the user, the content feature vector of the watching content and the score, and sending the score and the updated interest feature vector of the user to the cloud server;
the cloud server is used for determining a user cluster which is closest to the user in all user clusters of the cloud server by using the received updated interest characteristic vector of the user when the network content needs to be recommended to the user, and adding the content corresponding to the user cluster into an initial recommendation queue; the intelligent terminal is also used for deleting the content of the behavior generated by the user in the initial recommendation queue, sequencing the updated initial recommendation queue and recommending the first N contents to the user through the intelligent terminal; wherein, N is a preset positive integer.
11. The recommendation system according to claim 10, wherein the cloud server comprises: the device comprises a receiving unit, a learning unit, a storage unit and a recommendation engine unit;
the receiving unit is used for receiving the score and the updated interest feature vector of the user;
the learning unit is used for updating the content feature vector of the watching content according to the score, the content feature vector of the watching content stored in the storage unit and the interest feature vector of the user before/after updating, and storing the updated content feature vector of the watching content in the storage unit;
the storage unit is used for storing interest feature vectors of all users and content feature vectors of all contents;
the recommendation engine unit is used for determining a user cluster which is closest to the user in all user clusters of the cloud server by using the received updated interest feature vector of the user when the network content needs to be recommended for the user, and adding the content corresponding to the user cluster into an initial recommendation queue; the cloud server is further used for determining a content cluster which is closest to the watching content in all content clusters of the cloud server by using the updated content feature vector of the watching content when the user is determined to be interested in the watching content according to the score, and adding the content corresponding to the content cluster into an initial recommendation queue; and the system is also used for deleting the content of the behavior generated by the user in the initial recommendation queue, sequencing the updated initial recommendation queue and recommending the first N contents to the user.
Priority Applications (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010524482.6A CN111694971A (en) | 2020-06-10 | 2020-06-10 | Content recommendation method and system |
| PCT/KR2021/095065 WO2021251806A1 (en) | 2020-06-10 | 2021-05-28 | Content recommendation method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010524482.6A CN111694971A (en) | 2020-06-10 | 2020-06-10 | Content recommendation method and system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111694971A true CN111694971A (en) | 2020-09-22 |
Family
ID=72480240
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010524482.6A Pending CN111694971A (en) | 2020-06-10 | 2020-06-10 | Content recommendation method and system |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN111694971A (en) |
| WO (1) | WO2021251806A1 (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112650946A (en) * | 2020-12-30 | 2021-04-13 | 中国联合网络通信集团有限公司 | Product information recommendation method, device and system and storage medium |
| CN115510317A (en) * | 2022-09-22 | 2022-12-23 | 常熟理工学院 | Cloud-edge collaborative interest point recommendation method and device |
| WO2023002161A1 (en) * | 2021-07-19 | 2023-01-26 | Raters Group Ltd | Apparatus for outputting content |
| CN116610868A (en) * | 2023-07-13 | 2023-08-18 | 支付宝(杭州)信息技术有限公司 | Sample labeling method, end-edge cloud cooperative training method and device |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115048504A (en) * | 2022-05-17 | 2022-09-13 | 咪咕文化科技有限公司 | Information pushing method and device, computer equipment and computer readable storage medium |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106844653A (en) * | 2017-01-20 | 2017-06-13 | 上海幻电信息科技有限公司 | A kind of media data recommends method and system |
| CN109543132A (en) * | 2018-11-22 | 2019-03-29 | 深圳墨世科技有限公司 | Content recommendation method, device, electronic equipment and storage medium |
| CN110070134A (en) * | 2019-04-25 | 2019-07-30 | 厦门快商通信息咨询有限公司 | A kind of recommended method and device based on user interest perception |
Family Cites Families (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP3997774B2 (en) * | 2001-12-11 | 2007-10-24 | ソニー株式会社 | Data processing system, data processing method, information processing apparatus, and computer program |
| JP5224868B2 (en) * | 2008-03-28 | 2013-07-03 | 株式会社東芝 | Information recommendation device and information recommendation method |
| EP2661749A1 (en) * | 2011-01-04 | 2013-11-13 | Thomson Licensing | Sequencing content |
| KR101573601B1 (en) * | 2014-03-10 | 2015-12-04 | 단국대학교 산학협력단 | Apparatus and method for hybrid filtering content recommendation using user profile and context information based on preference |
| JP7070093B2 (en) * | 2018-05-25 | 2022-05-18 | 日本電信電話株式会社 | Clustering device, clustering method and program |
-
2020
- 2020-06-10 CN CN202010524482.6A patent/CN111694971A/en active Pending
-
2021
- 2021-05-28 WO PCT/KR2021/095065 patent/WO2021251806A1/en active Application Filing
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106844653A (en) * | 2017-01-20 | 2017-06-13 | 上海幻电信息科技有限公司 | A kind of media data recommends method and system |
| CN109543132A (en) * | 2018-11-22 | 2019-03-29 | 深圳墨世科技有限公司 | Content recommendation method, device, electronic equipment and storage medium |
| CN110070134A (en) * | 2019-04-25 | 2019-07-30 | 厦门快商通信息咨询有限公司 | A kind of recommended method and device based on user interest perception |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112650946A (en) * | 2020-12-30 | 2021-04-13 | 中国联合网络通信集团有限公司 | Product information recommendation method, device and system and storage medium |
| CN112650946B (en) * | 2020-12-30 | 2023-06-09 | 中国联合网络通信集团有限公司 | Product information recommendation method, device, system and storage medium |
| WO2023002161A1 (en) * | 2021-07-19 | 2023-01-26 | Raters Group Ltd | Apparatus for outputting content |
| CN115510317A (en) * | 2022-09-22 | 2022-12-23 | 常熟理工学院 | Cloud-edge collaborative interest point recommendation method and device |
| CN116610868A (en) * | 2023-07-13 | 2023-08-18 | 支付宝(杭州)信息技术有限公司 | Sample labeling method, end-edge cloud cooperative training method and device |
| CN116610868B (en) * | 2023-07-13 | 2023-09-29 | 支付宝(杭州)信息技术有限公司 | Sample labeling method, end-edge cloud cooperative training method and device |
Also Published As
| Publication number | Publication date |
|---|---|
| WO2021251806A1 (en) | 2021-12-16 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111694971A (en) | Content recommendation method and system | |
| CN106940801B (en) | A kind of deeply study recommender system and method for Wide Area Network | |
| US20130283303A1 (en) | Apparatus and method for recommending content based on user's emotion | |
| US10306287B2 (en) | System and method for organizing multimedia content | |
| CN102056018B (en) | Method and system for providing TV guide and method for providing program-requesting information | |
| CN102955781B (en) | A kind of personage's searching method and device | |
| CN108595493B (en) | Media content pushing method and device, storage medium and electronic device | |
| CN104869439B (en) | A kind of video pushing method and device | |
| US10178365B1 (en) | System and method for combining audio tracks with video files | |
| CN106791964B (en) | Broadcast TV program recommender system and method | |
| CN103605723A (en) | Video recommending method based on particle swarm algorithm | |
| CN111143699B (en) | Recommendation system based on similarity and confidence coefficient clustering | |
| CN114140171A (en) | Robust unbiased negative sampling method applied to recommendation system | |
| CN111552831B (en) | Music recommendation method and server | |
| CN118035489A (en) | Video searching method and device, storage medium and electronic equipment | |
| CN111954017B (en) | Live broadcast room searching method and device, server and storage medium | |
| CN114332550A (en) | Model training method, system, storage medium and terminal equipment | |
| EP2741507B1 (en) | Video processing system, method of determining viewer preference, video processing apparatus, and control method and control program therefor | |
| CN114449342B (en) | Video recommendation method, device, computer readable storage medium and computer equipment | |
| CN102447737A (en) | Service push method based on cloud platform | |
| CN108804492B (en) | Method and device for recommending multimedia objects | |
| JP2008242689A (en) | Content classification system, server, terminal device, program, and recording medium | |
| CN114581139A (en) | Content updating method based on big data intention mining and deep learning service system | |
| CN114595277A (en) | Data mining method and system combining cloud computing and dynamic analysis | |
| CN109271580B (en) | Search method, device, client and search engine |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20200922 |
|
| RJ01 | Rejection of invention patent application after publication |