CN-122002089-A - Server and sample data generation method

CN122002089ACN 122002089 ACN122002089 ACN 122002089ACN-122002089-A

Abstract

The application relates to a server and a sample data generation method, and relates to the technical field of artificial intelligence. The server comprises a first communication device, at least one processor, a first communication device, a second communication device, a first processor, a second communication device, a third communication device, a fourth communication device, a fifth communication device, a sixth communication device, a seventh communication device and a fourth communication device, wherein the first communication device is in communication connection with the terminal, the at least one processor is connected with the first communication device and is configured to respond to a media asset recommendation request of the terminal, generate an association identifier according to a receiving timestamp and a target user identifier, acquire a current user characteristic from a characteristic library according to the target user identifier, call an initial recommendation model, select target media assets from candidate media assets according to the current user characteristic and the current media asset characteristic, associate the association identifier, the current user characteristic and the current media asset characteristic, send recommendation information to the terminal, receive user log data of the terminal, and construct sample data for training the initial recommendation model based on the current user characteristic, response behavior data and the current media asset characteristic associated with the association identifier under the condition that sample construction conditions are met. The accuracy of the input content of the recommendation model is improved.

Inventors

You Shukai
HUANG SHANSHAN
RAO GANG
WANG BAOYUN

Assignees

青岛聚看云科技有限公司

Dates

Publication Date: 20260508
Application Date: 20260129

Claims (10)

1. A server for a server, which comprises a server and a server, characterized by comprising the following steps: a first communication device configured to be in communication connection with a terminal; and at least one processor coupled to the first communication device and configured to: responding to a media resource recommendation request of a terminal, and generating an association identifier corresponding to the media resource recommendation request according to a receiving time stamp of the media resource recommendation request and a target user identifier carried by the media resource recommendation request; Acquiring current user characteristics of a target user from a characteristic library according to the target user identification, wherein the characteristic library stores user characteristics corresponding to different user identifications; calling an initial recommendation model, and selecting a target media asset from the candidate media assets according to the current user characteristics and the current media asset characteristics of the candidate media assets; associating the associated identifier, the current user characteristic and the current media asset characteristic of the target media asset, and The recommendation information is sent to the terminal, wherein the recommendation information comprises the association identifier and the media information of the target media; Receiving user log data of the terminal, wherein the user log data comprises response behavior data associated with the association identifier, and the response behavior data is behavior data of the target user aiming at the target media asset; and under the condition that the sample construction condition is met, constructing sample data for training the initial recommendation model based on the current user characteristic, the response behavior data and the current media resource characteristic of the target media resource which are associated with the association identifier.
2. The server of claim 1, wherein the processor is configured to, when executing the generation of the association identifier corresponding to the media asset recommendation request according to the receiving timestamp of the media asset recommendation request and the target user identifier carried by the media asset recommendation request: and splicing the receiving time stamp of the media asset recommendation request, the target user identification carried by the media asset recommendation request and the random number to obtain an associated identification corresponding to the media asset recommendation request.
3. The server of claim 1, wherein the processor, when executing sending recommendation information to the terminal, is configured to: and adding the association identifier into a preset field of the media information of the target media information to obtain recommendation information, and sending the recommendation information to the terminal.
4. The server of any of claims 1-3, wherein the processor, when executing the current asset characteristics based on the current user characteristic, the response behavior data, and the target asset associated with the association identifier, is configured to, when constructing sample data for training the initial recommendation model: Acquiring a first time stamp and a second time stamp, wherein the first time stamp is a time stamp for acquiring a current feature, and the second time stamp is a time stamp recorded in the user log data for generating the response behavior data; verifying the timeliness between the current feature and the response behavior data according to the first timestamp and the second timestamp; Determining a time difference between the first timestamp and the second timestamp if the timing verification passes; And under the condition that the time difference value is smaller than a time threshold value, constructing sample data for training the initial recommendation model based on the current user characteristic, the response behavior data and the current media resource characteristic of the target media resource which are associated with the association identifier.
5. The server of claim 4, wherein the processor, when executing the checking of the timeliness between the current characteristic and the response behavior data according to the first timestamp and the second timestamp, is configured to: comparing the first timestamp with the second timestamp; Determining that a time-sequential check between the current feature and the response behavior data passes if the first timestamp is less than the second timestamp; And in the case that the first time stamp is not smaller than the second time stamp, determining that the time sequence check between the current feature and the response behavior data is not passed.
6. The server of any of claims 1-3, wherein the media asset recommendation request further includes a scene identifier and a column identifier, wherein the processor, when executing the call initial recommendation model to select a target media asset from the candidate media assets based on the current user characteristic and the current media asset characteristic of the candidate media asset, is configured to: Selecting an initial recommendation model corresponding to the column identification from different candidate recommendation models according to the column identification; Invoking the initial recommendation model, matching a target feature with a current media asset feature of a candidate media asset, and selecting a target media asset from the candidate media asset according to a matching result, wherein the target feature comprises the current user feature and the scene identifier; The processor, executing the sample data for training the initial recommendation model based on the current user characteristics associated with the associated identification, the response behavior data, and the current asset characteristics of the target asset, is configured to: sample data for training the initial recommendation model is constructed based on the current user characteristics associated with the association identifier, the response behavior data, the current media asset characteristics of the target media asset, and the scene identifier.
7. The server of any of claims 1-3, wherein the processor, when executing the construction of sample data for training the initial recommendation model based on the current user characteristics associated with the associated identification, the response behavior data, and the current asset characteristics of the target asset, is configured to: If the response behavior data is first response behavior data, positive sample data for training the initial recommendation model is constructed based on the current user characteristics associated with the association identifier and the current media asset characteristics of the target media asset, wherein the first response behavior data comprises behavior data for characterizing that the target user pays attention to the target media asset; And if the response behavior data is second response behavior data, constructing negative sample data for training the initial recommendation model based on the current user characteristics associated with the association identifier and the current media asset characteristics of the target media asset, wherein the second response behavior data comprises behavior data representing that the target user does not pay attention to the target media asset.
8. A server according to any of claims 1-3, wherein the sample construction conditions comprise any of the following: The current residual computing resources of the server are larger than or equal to the computing resources required for constructing the sample data; The current time is in a preset time period.
9. A terminal, comprising: A display; A second communication device configured to be communicatively connected to the server; And at least one controller connected to the second communication device and configured to: Sending a media asset recommendation request to the server; The method comprises the steps of receiving recommendation information sent by a server, wherein the recommendation information comprises an association identifier and media asset information of target media assets, the association identifier is generated by the server based on a receiving timestamp of a media asset recommendation request and a target user identifier carried by the media asset recommendation request; Controlling the display to display the media information, and acquiring response behavior data of the target user, which is fed back based on the media information, aiming at the target media information; Associating the association identifier with the response behavior data and generating user log data comprising the response behavior data; And the user log data are used for the server to construct sample data used for training the initial recommendation model based on the current user characteristic, the response behavior data and the current media resource characteristic of the target media resource which are associated with the association identifier under the condition that the sample construction condition is met.
10. A sample data generation method, applied to a server, comprising: responding to a media resource recommendation request of a terminal, and generating an association identifier corresponding to the media resource recommendation request according to a receiving time stamp of the media resource recommendation request and a target user identifier carried by the media resource recommendation request; Acquiring current user characteristics of a target user from a characteristic library according to the target user identification, wherein the characteristic library stores user characteristics corresponding to different user identifications; calling an initial recommendation model, and selecting a target media asset from the candidate media assets according to the current user characteristics and the current media asset characteristics of the candidate media assets; associating the associated identifier, the current user characteristic and the current media asset characteristic of the target media asset, and The recommendation information is sent to the terminal, wherein the recommendation information comprises the association identifier and the media information of the target media; Receiving user log data of the terminal, wherein the user log data comprises response behavior data associated with the association identifier, and the response behavior data is behavior data of the target user aiming at the target media asset; and under the condition that the sample construction condition is met, constructing sample data for training the initial recommendation model based on the current user characteristic, the response behavior data and the current media resource characteristic of the target media resource which are associated with the association identifier.

Description

Server and sample data generation method Technical Field The application relates to the technical field of artificial intelligence, in particular to a server and a sample data generation method. Background Over-the-Top (OTT) video intelligent recommendation is based on OTT terminals, such as large-screen devices of intelligent televisions, set-Top boxes and the like, user data is analyzed through a recommendation model, and personalized video content, such as episodes, movies, and process, is accurately pushed to users, so that the recommended content is more fit with individual preferences of the users. In the related art, the sample data used for training the recommendation model cannot accurately reflect the related characteristics of the recommendation scene, so that the content output by the recommendation model cannot be well attached to the individual preference of the user. Disclosure of Invention The application provides a server and a sample data generation method, which can enable sample data used for training a recommendation model to accurately reflect relevant characteristics of a recommendation scene, so that content output by the recommendation model can be better attached to individual preferences of users. In a first aspect, some embodiments provide a server comprising: a first communication device configured to be in communication connection with a terminal; and at least one processor coupled to the first communication device and configured to: responding to a media resource recommendation request of a terminal, and generating an association identifier corresponding to the media resource recommendation request according to a receiving time stamp of the media resource recommendation request and a target user identifier carried by the media resource recommendation request; Acquiring current user characteristics of a target user from a characteristic library according to the target user identification, wherein the characteristic library stores user characteristics corresponding to different user identifications; calling an initial recommendation model, and selecting a target media asset from the candidate media assets according to the current user characteristics and the current media asset characteristics of the candidate media assets; associating the associated identifier, the current user characteristic and the current media asset characteristic of the target media asset, and The recommendation information is sent to the terminal, wherein the recommendation information comprises the association identifier and the media information of the target media; Receiving user log data of the terminal, wherein the user log data comprises response behavior data associated with the association identifier, and the response behavior data is behavior data of the target user aiming at the target media asset; and under the condition that the sample construction condition is met, constructing sample data for training the initial recommendation model based on the current user characteristic, the response behavior data and the current media resource characteristic of the target media resource which are associated with the association identifier. In the above embodiment, the server generates the association identifier by responding to the media asset recommendation request of the terminal, associates the association identifier with the current user feature and the current media asset feature of the target media asset, and constructs sample data by combining the response behavior data of the user for the target media asset. Because the sample data contains the real-time user characteristics and the real-time media resource characteristics at the reasoning moment and the real user response behaviors corresponding to the characteristics, the sample data can accurately reflect the related characteristics of the recommended scene, and the recommended model learns the mapping relation between the characteristics and the behaviors of the real recommended scene more closely under the training of the sample data, so that the accuracy of the recommended model is improved, and the recommended content more conforming to the individual preferences of the user is output by the recommended model. In a second aspect, some embodiments further provide a terminal, including: A display; A second communication device configured to be communicatively connected to the server; And at least one controller connected to the second communication device and configured to: Sending a media asset recommendation request to the server; The method comprises the steps of receiving recommendation information sent by a server, wherein the recommendation information comprises an association identifier and media asset information of target media assets, the association identifier is generated by the server based on a receiving timestamp of a media asset recommendation request and a target user identifier carried by the media asset recommendation req